diff options
| author | Jules Laplace <julescarbon@gmail.com> | 2018-11-25 22:19:15 +0100 |
|---|---|---|
| committer | Jules Laplace <julescarbon@gmail.com> | 2018-11-25 22:19:15 +0100 |
| commit | ee3d0d98e19f1d8177d85af1866fd0ee431fe9ea (patch) | |
| tree | 41372528e78d4328bc2a47bbbabac7e809c58894 /scraper/reports/institutions_missing.html | |
| parent | 255b8178af1e25a71fd23703d30c0d1f74911f47 (diff) | |
moving stuff
Diffstat (limited to 'scraper/reports/institutions_missing.html')
| -rw-r--r-- | scraper/reports/institutions_missing.html | 11693 |
1 files changed, 11693 insertions, 0 deletions
diff --git a/scraper/reports/institutions_missing.html b/scraper/reports/institutions_missing.html new file mode 100644 index 00000000..93a26238 --- /dev/null +++ b/scraper/reports/institutions_missing.html @@ -0,0 +1,11693 @@ +<!doctype html><html><head><title>Institutions</title><link rel='stylesheet' href='reports.css'></head><body><h2>Institutions</h2><table border='1' cellpadding='3' cellspacing='3'><tr><td>61084a25ebe736e8f6d7a6e53b2c20d9723c4608</td><td></td></tr><tr><td>61f04606528ecf4a42b49e8ac2add2e9f92c0def</td><td>Deep Deformation Network for Object Landmark +<br/>Localization +<br/>NEC Laboratories America, Department of Media Analytics +</td></tr><tr><td>614a7c42aae8946c7ad4c36b53290860f6256441</td><td>1 +<br/>Joint Face Detection and Alignment using +<br/>Multi-task Cascaded Convolutional Networks +</td></tr><tr><td>0d88ab0250748410a1bc990b67ab2efb370ade5d</td><td>Author(s) : +<br/>ERROR HANDLING IN MULTIMODAL BIOMETRIC SYSTEMS USING +<br/>RELIABILITY MEASURES (ThuPmOR6) +<br/>(EPFL, Switzerland) +<br/>(EPFL, Switzerland) +<br/>(EPFL, Switzerland) +<br/>(EPFL, Switzerland) +<br/>Plamen Prodanov +</td></tr><tr><td>0d467adaf936b112f570970c5210bdb3c626a717</td><td></td></tr><tr><td>0d6b28691e1aa2a17ffaa98b9b38ac3140fb3306</td><td>Review of Perceptual Resemblance of Local +<br/>Plastic Surgery Facial Images using Near Sets +<br/>1,2 Department of Computer Technology, +<br/>YCCE Nagpur, India +</td></tr><tr><td>0db8e6eb861ed9a70305c1839eaef34f2c85bbaf</td><td></td></tr><tr><td>0dbf4232fcbd52eb4599dc0760b18fcc1e9546e9</td><td></td></tr><tr><td>0d760e7d762fa449737ad51431f3ff938d6803fe</td><td>LCDet: Low-Complexity Fully-Convolutional Neural Networks for +<br/>Object Detection in Embedded Systems +<br/>UC San Diego ∗ +<br/>Gokce Dane +<br/>Qualcomm Inc. +<br/>UC San Diego +<br/>Qualcomm Inc. +<br/>UC San Diego +</td></tr><tr><td>0dd72887465046b0f8fc655793c6eaaac9c03a3d</td><td>Real-time Head Orientation from a Monocular +<br/>Camera using Deep Neural Network +<br/>KAIST, Republic of Korea +</td></tr><tr><td>0d087aaa6e2753099789cd9943495fbbd08437c0</td><td></td></tr><tr><td>0d8415a56660d3969449e77095be46ef0254a448</td><td></td></tr><tr><td>0d735e7552af0d1dcd856a8740401916e54b7eee</td><td></td></tr><tr><td>0d06b3a4132d8a2effed115a89617e0a702c957a</td><td></td></tr><tr><td>0d2dd4fc016cb6a517d8fb43a7cc3ff62964832e</td><td></td></tr><tr><td>0d33b6c8b4d1a3cb6d669b4b8c11c2a54c203d1a</td><td>Detection and Tracking of Faces in Videos: A Review +<br/>© 2016 IJEDR | Volume 4, Issue 2 | ISSN: 2321-9939 +<br/>of Related Work +<br/>1Student, 2Assistant Professor +<br/>1, 2Dept. of Electronics & Comm., S S I E T, Punjab, India +<br/>________________________________________________________________________________________________________ +</td></tr><tr><td>956317de62bd3024d4ea5a62effe8d6623a64e53</td><td>Lighting Analysis and Texture Modification of 3D Human +<br/>Face Scans +<br/>Author +<br/>Zhang, Paul, Zhao, Sanqiang, Gao, Yongsheng +<br/>Published +<br/>2007 +<br/>Conference Title +<br/>Digital Image Computing Techniques and Applications +<br/>DOI +<br/>https://doi.org/10.1109/DICTA.2007.4426825 +<br/>Copyright Statement +<br/>© 2007 IEEE. Personal use of this material is permitted. However, permission to reprint/ +<br/>republish this material for advertising or promotional purposes or for creating new collective +<br/>works for resale or redistribution to servers or lists, or to reuse any copyrighted component of +<br/>this work in other works must be obtained from the IEEE. +<br/>Downloaded from +<br/>http://hdl.handle.net/10072/17889 +<br/>Link to published version +<br/>http://www.ieee.org/ +<br/>Griffith Research Online +<br/>https://research-repository.griffith.edu.au +</td></tr><tr><td>956c634343e49319a5e3cba4f2bd2360bdcbc075</td><td>IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART B: CYBERNETICS, VOL. 36, NO. 4, AUGUST 2006 +<br/>873 +<br/>A Novel Incremental Principal Component Analysis +<br/>and Its Application for Face Recognition +</td></tr><tr><td>958c599a6f01678513849637bec5dc5dba592394</td><td>Noname manuscript No. +<br/>(will be inserted by the editor) +<br/>Generalized Zero-Shot Learning for Action +<br/>Recognition with Web-Scale Video Data +<br/>Received: date / Accepted: date +</td></tr><tr><td>59fc69b3bc4759eef1347161e1248e886702f8f7</td><td>Final Report of Final Year Project +<br/>HKU-Face: A Large Scale Dataset for +<br/>Deep Face Recognition +<br/>3035141841 +<br/>COMP4801 Final Year Project +<br/>Project Code: 17007 +</td></tr><tr><td>59bfeac0635d3f1f4891106ae0262b81841b06e4</td><td>Face Verification Using the LARK Face +<br/>Representation +</td></tr><tr><td>590628a9584e500f3e7f349ba7e2046c8c273fcf</td><td></td></tr><tr><td>59eefa01c067a33a0b9bad31c882e2710748ea24</td><td>IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY +<br/>Fast Landmark Localization +<br/>with 3D Component Reconstruction and CNN for +<br/>Cross-Pose Recognition +</td></tr><tr><td>5945464d47549e8dcaec37ad41471aa70001907f</td><td>Noname manuscript No. +<br/>(will be inserted by the editor) +<br/>Every Moment Counts: Dense Detailed Labeling of Actions in Complex +<br/>Videos +<br/>Received: date / Accepted: date +</td></tr><tr><td>59c9d416f7b3d33141cc94567925a447d0662d80</td><td>Universität des Saarlandes +<br/>Max-Planck-Institut für Informatik +<br/>AG5 +<br/>Matrix factorization over max-times +<br/>algebra for data mining +<br/>Masterarbeit im Fach Informatik +<br/>Master’s Thesis in Computer Science +<br/>von / by +<br/>angefertigt unter der Leitung von / supervised by +<br/>begutachtet von / reviewers +<br/>November 2013 +<br/>UNIVERSITASSARAVIENSIS </td></tr><tr><td>59a35b63cf845ebf0ba31c290423e24eb822d245</td><td>The FaceSketchID System: Matching Facial +<br/>Composites to Mugshots +<br/>tedious, and may not +</td></tr><tr><td>59f325e63f21b95d2b4e2700c461f0136aecc171</td><td>3070 +<br/>978-1-4577-1302-6/11/$26.00 ©2011 IEEE +<br/>FOR FACE RECOGNITION +<br/>1. INTRODUCTION +</td></tr><tr><td>5922e26c9eaaee92d1d70eae36275bb226ecdb2e</td><td>Boosting Classification Based Similarity +<br/>Learning by using Standard Distances +<br/>Departament d’Informàtica, Universitat de València +<br/>Av. de la Universitat s/n. 46100-Burjassot (Spain) +</td></tr><tr><td>59031a35b0727925f8c47c3b2194224323489d68</td><td>Sparse Variation Dictionary Learning for Face Recognition with A Single +<br/>Training Sample Per Person +<br/>ETH Zurich +<br/>Switzerland +</td></tr><tr><td>926c67a611824bc5ba67db11db9c05626e79de96</td><td>1913 +<br/>Enhancing Bilinear Subspace Learning +<br/>by Element Rearrangement +</td></tr><tr><td>923ede53b0842619831e94c7150e0fc4104e62f7</td><td>978-1-4799-9988-0/16/$31.00 ©2016 IEEE +<br/>1293 +<br/>ICASSP 2016 +</td></tr><tr><td>92b61b09d2eed4937058d0f9494d9efeddc39002</td><td>Under review in IJCV manuscript No. +<br/>(will be inserted by the editor) +<br/>BoxCars: Improving Vehicle Fine-Grained Recognition using +<br/>3D Bounding Boxes in Traffic Surveillance +<br/>Received: date / Accepted: date +</td></tr><tr><td>920a92900fbff22fdaaef4b128ca3ca8e8d54c3e</td><td>LEARNING PATTERN TRANSFORMATION MANIFOLDS WITH PARAMETRIC ATOM +<br/>SELECTION +<br/>Ecole Polytechnique F´ed´erale de Lausanne (EPFL) +<br/>Signal Processing Laboratory (LTS4) +<br/>Switzerland-1015 Lausanne +</td></tr><tr><td>9207671d9e2b668c065e06d9f58f597601039e5e</td><td>Face Detection Using a 3D Model on +<br/>Face Keypoints +</td></tr><tr><td>9282239846d79a29392aa71fc24880651826af72</td><td>Antonakos et al. EURASIP Journal on Image and Video Processing 2014, 2014:14 +<br/>http://jivp.eurasipjournals.com/content/2014/1/14 +<br/>RESEARCH +<br/>Open Access +<br/>Classification of extreme facial events in sign +<br/>language videos +</td></tr><tr><td>92c2dd6b3ac9227fce0a960093ca30678bceb364</td><td>Provided by the author(s) and NUI Galway in accordance with publisher policies. Please cite the published +<br/>version when available. +<br/>Title +<br/>On color texture normalization for active appearance models +<br/>Author(s) +<br/>Ionita, Mircea C.; Corcoran, Peter M.; Buzuloiu, Vasile +<br/>Publication +<br/>Date +<br/>2009-05-12 +<br/>Publication +<br/>Information +<br/>Ionita, M. C., Corcoran, P., & Buzuloiu, V. (2009). On Color +<br/>Texture Normalization for Active Appearance Models. Image +<br/>Processing, IEEE Transactions on, 18(6), 1372-1378. +<br/>Publisher +<br/>IEEE +<br/>Link to +<br/>publisher's +<br/>version +<br/>http://dx.doi.org/10.1109/TIP.2009.2017163 +<br/>Item record +<br/>http://hdl.handle.net/10379/1350 +<br/>Some rights reserved. For more information, please see the item record link above. +<br/>Downloaded 2018-11-06T00:40:53Z +</td></tr><tr><td>92fada7564d572b72fd3be09ea3c39373df3e27c</td><td></td></tr><tr><td>927ad0dceacce2bb482b96f42f2fe2ad1873f37a</td><td>Interest-Point based Face Recognition System +<br/>87 +<br/>X +<br/>Interest-Point based Face Recognition System +<br/>Spain +<br/>1. Introduction +<br/>Among all applications of face recognition systems, surveillance is one of the most +<br/>challenging ones. In such an application, the goal is to detect known criminals in crowded +<br/>environments, like airports or train stations. Some attempts have been made, like those of +<br/>Tokio (Engadget, 2006) or Mainz (Deutsche Welle, 2006), with limited success. +<br/>The first task to be carried out in an automatic surveillance system involves the detection of +<br/>all the faces in the images taken by the video cameras. Current face detection algorithms are +<br/>highly reliable and thus, they will not be the focus of our work. Some of the best performing +<br/>examples are the Viola-Jones algorithm (Viola & Jones, 2004) or the Schneiderman-Kanade +<br/>algorithm (Schneiderman & Kanade, 2000). +<br/>The second task to be carried out involves the comparison of all detected faces among the +<br/>database of known criminals. The ideal behaviour of an automatic system performing this +<br/>task would be to get a 100% correct identification rate, but this behaviour is far from the +<br/>capabilities of current face recognition algorithms. Assuming that there will be false +<br/>identifications, supervised surveillance systems seem to be the most realistic option: the +<br/>automatic system issues an alarm whenever it detects a possible match with a criminal, and +<br/>a human decides whether it is a false alarm or not. Figure 1 shows an example. +<br/>However, even in a supervised scenario the requirements for the face recognition algorithm +<br/>are extremely high: the false alarm rate must be low enough as to allow the human operator +<br/>to cope with it; and the percentage of undetected criminals must be kept to a minimum in +<br/>order to ensure security. Fulfilling both requirements at the same time is the main challenge, +<br/>as a reduction in false alarm rate usually implies an increase of the percentage of undetected +<br/>criminals. +<br/>We propose a novel face recognition system based in the use of interest point detectors and +<br/>local descriptors. In order to check the performances of our system, and particularly its +<br/>performances in a surveillance application, we present experimental results in terms of +<br/>Receiver Operating Characteristic curves or ROC curves. From the experimental results, it +<br/>becomes clear that our system outperforms classical appearance based approaches. +<br/>www.intechopen.com +</td></tr><tr><td>929bd1d11d4f9cbc638779fbaf958f0efb82e603</td><td>This is the author’s version of a work that was submitted/accepted for pub- +<br/>lication in the following source: +<br/>Zhang, Ligang & Tjondronegoro, Dian W. (2010) Improving the perfor- +<br/>mance of facial expression recognition using dynamic, subtle and regional +<br/>features. +<br/>In Kok, WaiWong, B. Sumudu, U. Mendis, & Abdesselam , +<br/>Bouzerdoum (Eds.) Neural Information Processing. Models and Applica- +<br/>tions, Lecture Notes in Computer Science, Sydney, N.S.W, pp. 582-589. +<br/>This file was downloaded from: http://eprints.qut.edu.au/43788/ +<br/>c(cid:13) Copyright 2010 Springer-Verlag +<br/>Conference proceedings published, by Springer Verlag, will be available +<br/>via Lecture Notes in Computer Science http://www.springer.de/comp/lncs/ +<br/>Notice: Changes introduced as a result of publishing processes such as +<br/>copy-editing and formatting may not be reflected in this document. For a +<br/>definitive version of this work, please refer to the published source: +<br/>http://dx.doi.org/10.1007/978-3-642-17534-3_72 +</td></tr><tr><td>0c36c988acc9ec239953ff1b3931799af388ef70</td><td>Face Detection Using Improved Faster RCNN +<br/>Huawei Cloud BU, China +<br/>Figure1.Face detection results of FDNet1.0 +</td></tr><tr><td>0c5ddfa02982dcad47704888b271997c4de0674b</td><td></td></tr><tr><td>0cccf576050f493c8b8fec9ee0238277c0cfd69a</td><td></td></tr><tr><td>0c069a870367b54dd06d0da63b1e3a900a257298</td><td>Author manuscript, published in "ICANN 2011 - International Conference on Artificial Neural Networks (2011)" +</td></tr><tr><td>0c75c7c54eec85e962b1720755381cdca3f57dfb</td><td>2212 +<br/>Face Landmark Fitting via Optimized Part +<br/>Mixtures and Cascaded Deformable Model +</td></tr><tr><td>0ca36ecaf4015ca4095e07f0302d28a5d9424254</td><td>Improving Bag-of-Visual-Words Towards Effective Facial Expressive +<br/>Image Classification +<br/>1Univ. Grenoble Alpes, CNRS, Grenoble INP∗ , GIPSA-lab, 38000 Grenoble, France +<br/>Keywords: +<br/>BoVW, k-means++, Relative Conjunction Matrix, SIFT, Spatial Pyramids, TF.IDF. +</td></tr><tr><td>0cfca73806f443188632266513bac6aaf6923fa8</td><td>Predictive Uncertainty in Large Scale Classification +<br/>using Dropout - Stochastic Gradient Hamiltonian +<br/>Monte Carlo. +<br/>Vergara, Diego∗1, Hern´andez, Sergio∗2, Valdenegro-Toro, Mat´ıas∗∗3 and Jorquera, Felipe∗4. +<br/>∗Laboratorio de Procesamiento de Informaci´on Geoespacial, Universidad Cat´olica del Maule, Chile. +<br/>∗∗German Research Centre for Artificial Intelligence, Bremen, Germany. +</td></tr><tr><td>0c54e9ac43d2d3bab1543c43ee137fc47b77276e</td><td></td></tr><tr><td>0c5afb209b647456e99ce42a6d9d177764f9a0dd</td><td>97 +<br/>Recognizing Action Units for +<br/>Facial Expression Analysis +</td></tr><tr><td>0c377fcbc3bbd35386b6ed4768beda7b5111eec6</td><td>258 +<br/>A Unified Probabilistic Framework +<br/>for Spontaneous Facial Action Modeling +<br/>and Understanding +</td></tr><tr><td>0cb2dd5f178e3a297a0c33068961018659d0f443</td><td></td></tr><tr><td>0cf7da0df64557a4774100f6fde898bc4a3c4840</td><td>Shape Matching and Object Recognition using Low Distortion Correspondences +<br/>Department of Electrical Engineering and Computer Science +<br/>U.C. Berkeley +</td></tr><tr><td>0c4659b35ec2518914da924e692deb37e96d6206</td><td>1236 +<br/>Registering a MultiSensor Ensemble of Images +</td></tr><tr><td>0c53ef79bb8e5ba4e6a8ebad6d453ecf3672926d</td><td>SUBMITTED TO JOURNAL +<br/>Weakly Supervised PatchNets: Describing and +<br/>Aggregating Local Patches for Scene Recognition +</td></tr><tr><td>0c60eebe10b56dbffe66bb3812793dd514865935</td><td></td></tr><tr><td>6601a0906e503a6221d2e0f2ca8c3f544a4adab7</td><td>SRTM-2 2/9/06 3:27 PM Page 321 +<br/>Detection of Ancient Settlement Mounds: +<br/>Archaeological Survey Based on the +<br/>SRTM Terrain Model +<br/>B.H. Menze, J.A. Ur, and A.G. Sherratt +</td></tr><tr><td>660b73b0f39d4e644bf13a1745d6ee74424d4a16</td><td></td></tr><tr><td>66d512342355fb77a4450decc89977efe7e55fa2</td><td>Under review as a conference paper at ICLR 2018 +<br/>LEARNING NON-LINEAR TRANSFORM WITH DISCRIM- +<br/>INATIVE AND MINIMUM INFORMATION LOSS PRIORS +<br/>Anonymous authors +<br/>Paper under double-blind review +</td></tr><tr><td>6643a7feebd0479916d94fb9186e403a4e5f7cbf</td><td>Chapter 8 +<br/>3D Face Recognition +</td></tr><tr><td>661ca4bbb49bb496f56311e9d4263dfac8eb96e9</td><td>Datasheets for Datasets +</td></tr><tr><td>66d087f3dd2e19ffe340c26ef17efe0062a59290</td><td>Dog Breed Identification +<br/>Brian Mittl +<br/>Vijay Singh +</td></tr><tr><td>66a2c229ac82e38f1b7c77a786d8cf0d7e369598</td><td>Proceedings of the 2016 Industrial and Systems Engineering Research Conference +<br/>H. Yang, Z. Kong, and MD Sarder, eds. +<br/>A Probabilistic Adaptive Search System +<br/>for Exploring the Face Space +<br/>Escuela Superior Politecnica del Litoral (ESPOL) +<br/>Guayaquil-Ecuador +</td></tr><tr><td>66886997988358847615375ba7d6e9eb0f1bb27f</td><td></td></tr><tr><td>66837add89caffd9c91430820f49adb5d3f40930</td><td></td></tr><tr><td>66a9935e958a779a3a2267c85ecb69fbbb75b8dc</td><td>FAST AND ROBUST FIXED-RANK MATRIX RECOVERY +<br/>Fast and Robust Fixed-Rank Matrix +<br/>Recovery +<br/>Antonio Lopez +</td></tr><tr><td>66533107f9abdc7d1cb8f8795025fc7e78eb1122</td><td>Vi a +<br/>i a Whee +<br/>W y g Sgy Dae i iy g S g iz ad Ze ga Biey +<br/>y EECS AST 373 1 g Dg Y g G Taej 305 701 REA +<br/>z VR Cee ETR 161 ajg Dg Y g G Taej 305 350 REA +<br/>Abac +<br/>Thee exi he c eaive aciviy bewee a h +<br/>a beig ad ehabi +<br/>a eae ehabi +<br/>e ad ha he bee(cid:12) f ehabi +<br/> ch a ai +<br/>eadig i e f he eeia +<br/>fied +<br/>cf ad afey f a +<br/>a +<br/>bic a ye ARES ad i h a b +<br/>ieaci ech +<br/>ech +<br/>a +<br/>vi a +<br/>ecgizig he iive ad egaive eaig f he +<br/> e i efed he bai f chage f he facia +<br/>exei a d +<br/> e iei whi +<br/> e wih a beveage. F he eÆcie vi a +<br/>i ceig +<br/>c +<br/>ed e(cid:11)ec f he bic a. The vi a +<br/>wih e(cid:11)ecive iei eadig i ccef +<br/> eve a beveage f he e. +<br/>d ci +<br/>Whee +<br/>ai he e +<br/>ca i ey ad f ci i +<br/>ye ci f a weed whee +<br/>a ad ha +<br/>he whee +<br/>he bic a ad h ake ib +<br/>exiece f a e ad a b i he ae evi +<br/>e. +<br/> hi cae he e eed ieac wih +<br/>he bic a i cfab +<br/>Fig e 1: The whee +<br/>h a b ieaci ech +<br/>eve i ha bee eed ha ay diÆc +<br/>i h a bf ieaci i exiig ehabi +<br/>b. F exa +<br/>a ake a high cgiive +<br/>hyica +<br/>eaig jyick dexe +<br/>de +<br/>ai e eed ha he diÆc +<br/>ig ehabi +<br/>a +<br/>id a he begiig [4]. Theefe h a fied +<br/>h a b ieaci i e f eeia +<br/>i a whee +<br/> hi ae we cide he whee +<br/>bic ye ARES AST Rehabi +<br/>gieeig Sevice ye which we ae deve +<br/>a a evice bic ye f he diab +<br/>e +<br/>i e Fig. 1. Ag h a b ieaci ech +<br/>i e vi a +</td></tr><tr><td>66810438bfb52367e3f6f62c24f5bc127cf92e56</td><td>Face Recognition of Illumination Tolerance in 2D +<br/>Subspace Based on the Optimum Correlation +<br/>Filter +<br/>Xu Yi +<br/>Department of Information Engineering, Hunan Industry Polytechnic, Changsha, China +<br/>images will be tested to project +</td></tr><tr><td>66af2afd4c598c2841dbfd1053bf0c386579234e</td><td>Noname manuscript No. +<br/>(will be inserted by the editor) +<br/>Context Assisted Face Clustering Framework with +<br/>Human-in-the-Loop +<br/>Received: date / Accepted: date +</td></tr><tr><td>66e6f08873325d37e0ec20a4769ce881e04e964e</td><td>Int J Comput Vis (2014) 108:59–81 +<br/>DOI 10.1007/s11263-013-0695-z +<br/>The SUN Attribute Database: Beyond Categories for Deeper Scene +<br/>Understanding +<br/>Received: 27 February 2013 / Accepted: 28 December 2013 / Published online: 18 January 2014 +<br/>© Springer Science+Business Media New York 2014 +</td></tr><tr><td>661da40b838806a7effcb42d63a9624fcd684976</td><td>53 +<br/>An Illumination Invariant Accurate +<br/>Face Recognition with Down Scaling +<br/>of DCT Coefficients +<br/>Department of Computer Science and Engineering, Amity School of Engineering and Technology, New Delhi, India +<br/>In this paper, a novel approach for illumination normal- +<br/>ization under varying lighting conditions is presented. +<br/>Our approach utilizes the fact that discrete cosine trans- +<br/>form (DCT) low-frequency coefficients correspond to +<br/>illumination variations in a digital image. Under varying +<br/>illuminations, the images captured may have low con- +<br/>trast; initially we apply histogram equalization on these +<br/>for contrast stretching. Then the low-frequency DCT +<br/>coefficients are scaled down to compensate the illumi- +<br/>nation variations. The value of scaling down factor and +<br/>the number of low-frequency DCT coefficients, which +<br/>are to be rescaled, are obtained experimentally. The +<br/>classification is done using k−nearest neighbor classi- +<br/>fication and nearest mean classification on the images +<br/>obtained by inverse DCT on the processed coefficients. +<br/>The correlation coefficient and Euclidean distance ob- +<br/>tained using principal component analysis are used as +<br/>distance metrics in classification. We have tested our +<br/>face recognition method using Yale Face Database B. +<br/>The results show that our method performs without any +<br/>error (100% face recognition performance), even on the +<br/>most extreme illumination variations. There are different +<br/>schemes in the literature for illumination normalization +<br/>under varying lighting conditions, but no one is claimed +<br/>to give 100% recognition rate under all illumination +<br/>variations for this database. The proposed technique is +<br/>computationally efficient and can easily be implemented +<br/>for real time face recognition system. +<br/>Keywords: discrete cosine transform, correlation co- +<br/>efficient, face recognition, illumination normalization, +<br/>nearest neighbor classification +<br/>1. Introduction +<br/>Two-dimensional pattern classification plays a +<br/>crucial role in real-world applications. To build +<br/>high-performance surveillance or information +<br/>security systems, face recognition has been +<br/>known as the key application attracting enor- +<br/>mous researchers highlighting on related topics +<br/>[1,2]. Even though current machine recognition +<br/>systems have reached a certain level of matu- +<br/>rity, their success is limited by the real appli- +<br/>cations constraints, like pose, illumination and +<br/>expression. The FERET evaluation shows that +<br/>the performance of a face recognition system +<br/>decline seriously with the change of pose and +<br/>illumination conditions [31]. +<br/>To solve the variable illumination problem a +<br/>variety of approaches have been proposed [3, 7- +<br/>11, 26-29]. Early work in illumination invariant +<br/>face recognition focused on image representa- +<br/>tions that are mostly insensitive to changes in +<br/>illumination. There were approaches in which +<br/>the image representations and distance mea- +<br/>sures were evaluated on a tightly controlled face +<br/>database that varied the face pose, illumination, +<br/>and expression. The image representations in- +<br/>clude edge maps, 2D Gabor-like filters, first and +<br/>second derivatives of the gray-level image, and +<br/>the logarithmic transformations of the intensity +<br/>image along with these representations [4]. +<br/>The different approaches to solve the prob- +<br/>lem of illumination invariant face recognition +<br/>can be broadly classified into two main cate- +<br/>gories. The first category is named as passive +<br/>approach in which the visual spectrum images +<br/>are analyzed to overcome this problem. The +<br/>approaches belonging to other category named +<br/>active, attempt to overcome this problem by +<br/>employing active imaging techniques to obtain +<br/>face images captured in consistent illumina- +<br/>tion condition, or images of illumination invari- +<br/>ant modalities. There is a hierarchical catego- +<br/>rization of these two approaches. An exten- +<br/>sive review of both approaches is given in [5]. +</td></tr><tr><td>3edb0fa2d6b0f1984e8e2c523c558cb026b2a983</td><td>Automatic Age Estimation Based on +<br/>Facial Aging Patterns +</td></tr><tr><td>3ee7a8107a805370b296a53e355d111118e96b7c</td><td></td></tr><tr><td>3e4acf3f2d112fc6516abcdddbe9e17d839f5d9b</td><td>Deep Value Networks Learn to +<br/>Evaluate and Iteratively Refine Structured Outputs +</td></tr><tr><td>3ea8a6dc79d79319f7ad90d663558c664cf298d4</td><td></td></tr><tr><td>3e4f84ce00027723bdfdb21156c9003168bc1c80</td><td>1979 +<br/>© EURASIP, 2011 - ISSN 2076-1465 +<br/>19th European Signal Processing Conference (EUSIPCO 2011) +<br/>INTRODUCTION +</td></tr><tr><td>3e685704b140180d48142d1727080d2fb9e52163</td><td>Single Image Action Recognition by Predicting +<br/>Space-Time Saliency +</td></tr><tr><td>3e687d5ace90c407186602de1a7727167461194a</td><td>Photo Tagging by Collection-Aware People Recognition +<br/>UFF +<br/>UFF +<br/>Asla S´a +<br/>FGV +<br/>IMPA +</td></tr><tr><td>501096cca4d0b3d1ef407844642e39cd2ff86b37</td><td>Illumination Invariant Face Image +<br/>Representation using Quaternions +<br/>Dayron Rizo-Rodr´ıguez, Heydi M´endez-V´azquez, and Edel Garc´ıa-Reyes +<br/>Advanced Technologies Application Center. 7a # 21812 b/ 218 and 222, +<br/>Rpto. Siboney, Playa, P.C. 12200, La Habana, Cuba. +</td></tr><tr><td>501eda2d04b1db717b7834800d74dacb7df58f91</td><td></td></tr><tr><td>5083c6be0f8c85815ead5368882b584e4dfab4d1</td><td> Please do not quote. In press, Handbook of affective computing. New York, NY: Oxford +<br/>Automated Face Analysis for Affective Computing +</td></tr><tr><td>500b92578e4deff98ce20e6017124e6d2053b451</td><td></td></tr><tr><td>50ff21e595e0ebe51ae808a2da3b7940549f4035</td><td>IEEE TRANSACTIONS ON LATEX CLASS FILES, VOL. XX, NO. X, AUGUST 2017 +<br/>Age Group and Gender Estimation in the Wild with +<br/>Deep RoR Architecture +</td></tr><tr><td>5042b358705e8d8e8b0655d07f751be6a1565482</td><td>International Journal of +<br/>Emerging Research in Management &Technology +<br/>ISSN: 2278-9359 (Volume-4, Issue-8) +<br/>Research Article +<br/> August +<br/> 2015 +<br/>Review on Emotion Detection in Image +<br/>CSE & PCET, PTU HOD, CSE & PCET, PTU +<br/> Punjab, India Punj ab, India +</td></tr><tr><td>50e47857b11bfd3d420f6eafb155199f4b41f6d7</td><td>International Journal of Computer, Consumer and Control (IJ3C), Vol. 2, No.1 (2013) +<br/>3D Human Face Reconstruction Using a Hybrid of Photometric +<br/>Stereo and Independent Component Analysis +</td></tr><tr><td>50eb75dfece76ed9119ec543e04386dfc95dfd13</td><td>Learning Visual Entities and their Visual Attributes from Text Corpora +<br/>Dept. of Computer Science +<br/>K.U.Leuven, Belgium +<br/>Dept. of Computer Science +<br/>K.U.Leuven, Belgium +<br/>Dept. of Computer Science +<br/>K.U.Leuven, Belgium +</td></tr><tr><td>50a0930cb8cc353e15a5cb4d2f41b365675b5ebf</td><td></td></tr><tr><td>50d15cb17144344bb1879c0a5de7207471b9ff74</td><td>Divide, Share, and Conquer: Multi-task +<br/>Attribute Learning with Selective Sharing +</td></tr><tr><td>5028c0decfc8dd623c50b102424b93a8e9f2e390</td><td>Published as a conference paper at ICLR 2017 +<br/>REVISITING CLASSIFIER TWO-SAMPLE TESTS +<br/>1Facebook AI Research, 2WILLOW project team, Inria / ENS / CNRS +</td></tr><tr><td>505e55d0be8e48b30067fb132f05a91650666c41</td><td>A Model of Illumination Variation for Robust Face Recognition +<br/>Institut Eur´ecom +<br/>Multimedia Communications Department +<br/>BP 193, 06904 Sophia Antipolis Cedex, France +</td></tr><tr><td>680d662c30739521f5c4b76845cb341dce010735</td><td>Int J Comput Vis (2014) 108:82–96 +<br/>DOI 10.1007/s11263-014-0716-6 +<br/>Part and Attribute Discovery from Relative Annotations +<br/>Received: 25 February 2013 / Accepted: 14 March 2014 / Published online: 26 April 2014 +<br/>© Springer Science+Business Media New York 2014 +</td></tr><tr><td>68d2afd8c5c1c3a9bbda3dd209184e368e4376b9</td><td>Representation Learning by Rotating Your Faces +</td></tr><tr><td>68a3f12382003bc714c51c85fb6d0557dcb15467</td><td></td></tr><tr><td>68d08ed9470d973a54ef7806318d8894d87ba610</td><td>Drive Video Analysis for the Detection of Traffic Near-Miss Incidents +</td></tr><tr><td>68caf5d8ef325d7ea669f3fb76eac58e0170fff0</td><td></td></tr><tr><td>68d4056765c27fbcac233794857b7f5b8a6a82bf</td><td>Example-Based Face Shape Recovery Using the +<br/>Zenith Angle of the Surface Normal +<br/>Mario Castel´an1, Ana J. Almaz´an-Delf´ın2, Marco I. Ram´ırez-Sosa-Mor´an3, +<br/>and Luz A. Torres-M´endez1 +<br/>1 CINVESTAV Campus Saltillo, Ramos Arizpe 25900, Coahuila, M´exico +<br/>2 Universidad Veracruzana, Facultad de F´ısica e Inteligencia Artificial, Xalapa 91000, +<br/>3 ITESM, Campus Saltillo, Saltillo 25270, Coahuila, M´exico +<br/>Veracruz, M´exico +</td></tr><tr><td>684f5166d8147b59d9e0938d627beff8c9d208dd</td><td>IEEE TRANS. NNLS, JUNE 2017 +<br/>Discriminative Block-Diagonal Representation +<br/>Learning for Image Recognition +</td></tr><tr><td>68cf263a17862e4dd3547f7ecc863b2dc53320d8</td><td></td></tr><tr><td>68e9c837431f2ba59741b55004df60235e50994d</td><td>Detecting Faces Using Region-based Fully +<br/>Convolutional Networks +<br/>Tencent AI Lab, China +</td></tr><tr><td>687e17db5043661f8921fb86f215e9ca2264d4d2</td><td>A Robust Elastic and Partial Matching Metric for Face Recognition +<br/>Microsoft Corporate +<br/>One Microsoft Way, Redmond, WA 98052 +</td></tr><tr><td>688754568623f62032820546ae3b9ca458ed0870</td><td>bioRxiv preprint first posted online Sep. 27, 2016; +<br/>doi: +<br/>http://dx.doi.org/10.1101/077784 +<br/>. +<br/>The copyright holder for this preprint (which was not +<br/>peer-reviewed) is the author/funder. It is made available under a +<br/>CC-BY-NC-ND 4.0 International license +<br/>. +<br/>Resting high frequency heart rate variability is not associated with the +<br/>recognition of emotional facial expressions in healthy human adults. +<br/>1 Univ. Grenoble Alpes, LPNC, F-38040, Grenoble, France +<br/>2 CNRS, LPNC UMR 5105, F-38040, Grenoble, France +<br/>3 IPSY, Université Catholique de Louvain, Louvain-la-Neuve, Belgium +<br/>4 Fund for Scientific Research (FRS-FNRS), Brussels, Belgium +<br/>Correspondence concerning this article should be addressed to Brice Beffara, Office E250, Institut +<br/>de Recherches en Sciences Psychologiques, IPSY - Place du Cardinal Mercier, 10 bte L3.05.01 B-1348 +<br/>Author note +<br/>This study explores whether the myelinated vagal connection between the heart and the brain +<br/>is involved in emotion recognition. The Polyvagal theory postulates that the activity of the +<br/>myelinated vagus nerve underlies socio-emotional skills. It has been proposed that the perception +<br/>of emotions could be one of this skills dependent on heart-brain interactions. However, this +<br/>assumption was differently supported by diverging results suggesting that it could be related to +<br/>confounded factors. In the current study, we recorded the resting state vagal activity (reflected by +<br/>High Frequency Heart Rate Variability, HF-HRV) of 77 (68 suitable for analysis) healthy human +<br/>adults and measured their ability to identify dynamic emotional facial expressions. Results show +<br/>that HF-HRV is not related to the recognition of emotional facial expressions in healthy human +<br/>adults. We discuss this result in the frameworks of the polyvagal theory and the neurovisceral +<br/>integration model. +<br/>Keywords: HF-HRV; autonomic flexibility; emotion identification; dynamic EFEs; Polyvagal +<br/>theory; Neurovisceral integration model +<br/>Word count: 9810 +<br/>10 +<br/>11 +<br/>12 +<br/>13 +<br/>14 +<br/>15 +<br/>16 +<br/>17 +<br/>Introduction +<br/>The behavior of an animal is said social when involved in in- +<br/>teractions with other animals (Ward & Webster, 2016). These +<br/>interactions imply an exchange of information, signals, be- +<br/>tween at least two animals. In humans, the face is an efficient +<br/>communication channel, rapidly providing a high quantity of +<br/>information. Facial expressions thus play an important role +<br/>in the transmission of emotional information during social +<br/>interactions. The result of the communication is the combina- +<br/>tion of transmission from the sender and decoding from the +<br/>receiver (Jack & Schyns, 2015). As a consequence, the quality +<br/>of the interaction depends on the ability to both produce and +<br/>identify facial expressions. Emotions are therefore a core +<br/>feature of social bonding (Spoor & Kelly, 2004). Health +<br/>of individuals and groups depend on the quality of social +<br/>bonds in many animals (Boyer, Firat, & Leeuwen, 2015; S. L. +<br/>Brown & Brown, 2015; Neuberg, Kenrick, & Schaller, 2011), +<br/>18 +<br/>19 +<br/>20 +<br/>21 +<br/>22 +<br/>23 +<br/>24 +<br/>25 +<br/>26 +<br/>27 +<br/>28 +<br/>29 +<br/>30 +<br/>31 +<br/>32 +<br/>33 +<br/>34 +<br/>35 +<br/>especially in highly social species such as humans (Singer & +<br/>Klimecki, 2014). +<br/>The recognition of emotional signals produced by others is +<br/>not independent from its production by oneself (Niedenthal, +<br/>2007). The muscles of the face involved in the production of +<br/>a facial expressions are also activated during the perception of +<br/>the same facial expressions (Dimberg, Thunberg, & Elmehed, +<br/>2000). In other terms, the facial mimicry of the perceived +<br/>emotional facial expression (EFE) triggers its sensorimotor +<br/>simulation in the brain, which improves the recognition abili- +<br/>ties (Wood, Rychlowska, Korb, & Niedenthal, 2016). Beyond +<br/>that, the emotion can be seen as the body -including brain- +<br/>dynamic itself (Gallese & Caruana, 2016) which helps to un- +<br/>derstand why behavioral simulation is necessary to understand +<br/>the emotion. +<br/>The interplay between emotion production, emotion percep- +<br/>tion, social communication and body dynamics has been sum- +<br/>marized in the framework of the polyvagal theory (Porges, +</td></tr><tr><td>68f9cb5ee129e2b9477faf01181cd7e3099d1824</td><td>ALDA Algorithms for Online Feature Extraction +</td></tr><tr><td>68bf34e383092eb827dd6a61e9b362fcba36a83a</td><td></td></tr><tr><td>6889d649c6bbd9c0042fadec6c813f8e894ac6cc</td><td>Analysis of Robust Soft Learning Vector +<br/>Quantization and an application to Facial +<br/>Expression Recognition +</td></tr><tr><td>68c17aa1ecbff0787709be74d1d98d9efd78f410</td><td>International Journal of Optomechatronics, 6: 92–119, 2012 +<br/>Copyright # Taylor & Francis Group, LLC +<br/>ISSN: 1559-9612 print=1559-9620 online +<br/>DOI: 10.1080/15599612.2012.663463 +<br/>GENDER CLASSIFICATION FROM FACE IMAGES +<br/>USING MUTUAL INFORMATION AND FEATURE +<br/>FUSION +<br/>Department of Electrical Engineering and Advanced Mining Technology +<br/>Center, Universidad de Chile, Santiago, Chile +<br/>In this article we report a new method for gender classification from frontal face images +<br/>using feature selection based on mutual information and fusion of features extracted from +<br/>intensity, shape, texture, and from three different spatial scales. We compare the results of +<br/>three different mutual information measures: minimum redundancy and maximal relevance +<br/>(mRMR), normalized mutual information feature selection (NMIFS), and conditional +<br/>mutual information feature selection (CMIFS). We also show that by fusing features +<br/>extracted from six different methods we significantly improve the gender classification +<br/>results relative to those previously published, yielding 99.13% of the gender classification +<br/>rate on the FERET database. +<br/>Keywords: Feature fusion, feature selection, gender classification, mutual information, real-time gender +<br/>classification +<br/>1. INTRODUCTION +<br/>During the 90’s, one of the main issues addressed in the area of computer +<br/>vision was face detection. Many methods and applications were developed including +<br/>the face detection used in many digital cameras nowadays. Gender classification is +<br/>important in many possible applications including electronic marketing. Displays +<br/>at retail stores could show products and offers according to the person gender as +<br/>the person passes in front of a camera at the store. This is not a simple task since +<br/>faces are not rigid and depend on illumination, pose, gestures, facial expressions, +<br/>occlusions (glasses), and other facial features (makeup, beard). The high variability +<br/>in the appearance of the face directly affects their detection and classification. Auto- +<br/>matic classification of gender from face images has a wide range of possible applica- +<br/>tions, ranging from human-computer interaction to applications in real-time +<br/>electronic marketing in retail stores (Shan 2012; Bekios-Calfa et al. 2011; Chu +<br/>et al. 2010; Perez et al. 2010a). +<br/>Automatic gender classification has a wide range of possible applications for +<br/>improving human-machine interaction and face identification methods (Irick et al. +<br/>ing.uchile.cl +<br/>92 +</td></tr><tr><td>6888f3402039a36028d0a7e2c3df6db94f5cb9bb</td><td>Under review as a conference paper at ICLR 2018 +<br/>CLASSIFIER-TO-GENERATOR ATTACK: ESTIMATION +<br/>OF TRAINING DATA DISTRIBUTION FROM CLASSIFIER +<br/>Anonymous authors +<br/>Paper under double-blind review +</td></tr><tr><td>574751dbb53777101502419127ba8209562c4758</td><td></td></tr><tr><td>57b8b28f8748d998951b5a863ff1bfd7ca4ae6a5</td><td></td></tr><tr><td>57101b29680208cfedf041d13198299e2d396314</td><td></td></tr><tr><td>57893403f543db75d1f4e7355283bdca11f3ab1b</td><td></td></tr><tr><td>57f8e1f461ab25614f5fe51a83601710142f8e88</td><td>Region Selection for Robust Face Verification using UMACE Filters +<br/>Department of Electrical, Electronic and Systems Engineering, Faculty of Engineering, +<br/>Universiti Kebangsaan Malaysia, 43600 Bangi, Selangor, Malaysia. +<br/>In this paper, we investigate the verification performances of four subdivided face images with varying expressions. The +<br/>objective of this study is to evaluate which part of the face image is more tolerant to facial expression and still retains its personal +<br/>characteristics due to the variations of the image. The Unconstrained Minimum Average Correlation Energy (UMACE) filter is +<br/>implemented to perform the verification process because of its advantages such as shift–invariance, ability to trade-off between +<br/>discrimination and distortion tolerance, e.g. variations in pose, illumination and facial expression. The database obtained from the +<br/>facial expression database of Advanced Multimedia Processing (AMP) Lab at CMU is used in this study. Four equal +<br/>sizes of face regions i.e. bottom, top, left and right halves are used for the purpose of this study. The results show that the bottom +<br/>half of the face region gives the best performance in terms of the PSR values with zero false accepted rate (FAR) and zero false +<br/>rejection rate (FRR) compared to the other three regions. +<br/>1. Introduction +<br/>Face recognition is a well established field of research, +<br/>and a large number of algorithms have been proposed in the +<br/>literature. Various classifiers have been explored to improve +<br/>the accuracy of face classification. The basic approach is to +<br/>use distance-base methods which measure Euclidean distance +<br/>between any two vectors and then compare it with the preset +<br/>threshold. Neural Networks are often used as classifiers due +<br/>to their powerful generation ability [1]. Support Vector +<br/>Machines (SVM) have been applied with encouraging results +<br/>[2]. +<br/>In biometric applications, one of the important tasks is the +<br/>matching process between an individual biometrics against +<br/>the database that has been prepared during the enrolment +<br/>stage. For biometrics systems such as face authentication that +<br/>use images as personal characteristics, biometrics sensor +<br/>output and image pre-processing play an important role since +<br/>the quality of a biometric input can change significantly due +<br/>to illumination, noise and pose variations. Over the years, +<br/>researchers have studied the role of illumination variation, +<br/>pose variation, facial expression, and occlusions in affecting +<br/>the performance of face verification systems [3]. +<br/>The Minimum Average Correlation Energy (MACE) +<br/>filters have been reported to be an alternative solution to these +<br/>problems because of the advantages such as shift-invariance, +<br/>close-form expressions and distortion-tolerance. MACE +<br/>filters have been successfully applied in the field of automatic +<br/>target recognition as well as in biometric verification [3][4]. +<br/>Face and fingerprint verification using correlation filters have +<br/>been investigated in [5] and [6], respectively. Savvides et.al +<br/>performed face authentication and identification using +<br/>correlation filters based on illumination variation [7]. In the +<br/>process of implementing correlation filters, the number of +<br/>training images used depends on the level of distortions +<br/>applied to the images [5], [6]. +<br/>In this study, we investigate which part of a face image is +<br/>more tolerant to facial expression and retains its personal +<br/>characteristics for the verification process. Four subdivided +<br/>face images, i.e. bottom, top, left and right halves, with +<br/>varying expressions are investigated. By identifying only the +<br/>region of the face that gives the highest verification +<br/>performance, that region can be used instead of the full-face +<br/>to reduce storage requirements. +<br/>2. Unconstrained Minimum Average Correlation +<br/>Energy (UMACE) Filter +<br/>Correlation filter theory and the descriptions of the design +<br/>of the correlation filter can be found in a tutorial survey paper +<br/>[8]. According to [4][6], correlation filter evolves from +<br/>matched filters which are optimal for detecting a known +<br/>reference image in the presence of additive white Gaussian +<br/>noise. However, the detection rate of matched filters +<br/>decreases significantly due to even the small changes of scale, +<br/>rotation and pose of the reference image. +<br/>the pre-specified peak values +<br/>In an effort to solve this problem, the Synthetic +<br/>Discriminant Function (SDF) filter and the Equal Correlation +<br/>Peak SDF (ECP SDF) filter ware introduced which allowed +<br/>several training images to be represented by a single +<br/>correlation filter. SDF filter produces pre-specified values +<br/>called peak constraints. These peak values correspond to the +<br/>authentic class or impostor class when an image is tested. +<br/>However, +<br/>to +<br/>misclassifications when the sidelobes are larger than the +<br/>controlled values at the origin. +<br/>Savvides et.al developed +<br/>the Minimum Average +<br/>Correlation Energy (MACE) filters [5]. This filter reduces the +<br/>large sidelobes and produces a sharp peak when the test +<br/>image is from the same class as the images that have been +<br/>used to design the filter. There are two kinds of variants that +<br/>can be used in order to obtain a sharp peak when the test +<br/>image belongs to the authentic class. The first MACE filter +<br/>variant minimizes the average correlation energy of the +<br/>training images while constraining the correlation output at +<br/>the origin to a specific value for each of the training images. +<br/>The second MACE filter variant is the Unconstrained +<br/>Minimum Average Correlation Energy (UMACE) filter +<br/>which also minimizes the average correlation output while +<br/>maximizing the correlation output at the origin [4]. +<br/>lead +<br/>Proceedings of the International Conference onElectrical Engineering and InformaticsInstitut Teknologi Bandung, Indonesia June 17-19, 2007B-67ISBN 978-979-16338-0-2611 </td></tr><tr><td>57a1466c5985fe7594a91d46588d969007210581</td><td>A Taxonomy of Face-models for System Evaluation +<br/>Motivation and Data Types +<br/>Synthetic Data Types +<br/>Unverified – Have no underlying physical or +<br/>statistical basis +<br/>Physics -Based – Based on structure and +<br/>materials combined with the properties +<br/>formally modeled in physics. +<br/>Statistical – Use statistics from real +<br/>data/experiments to estimate/learn model +<br/>parameters. Generally have measurements +<br/>of accuracy +<br/>Guided Synthetic – Individual models based +<br/>on individual people. No attempt to capture +<br/>properties of large groups, a unique model +<br/>per person. For faces, guided models are +<br/>composed of 3D structure models and skin +<br/>textures, capturing many artifacts not +<br/>easily parameterized. Can be combined with +<br/>physics-based rendering to generate samples +<br/>under different conditions. +<br/>Semi–Synethetic – Use measured data such +<br/>as 2D images or 3D facial scans. These are +<br/>not truly synthetic as they are re-rendering’s +<br/>of real measured data. +<br/>Semi and Guided Synthetic data provide +<br/>higher operational relevance while +<br/>maintaining a high degree of control. +<br/>Generating statistically significant size +<br/>datasets for face matching system +<br/>evaluation is both a laborious and +<br/>expensive process. +<br/>There is a gap in datasets that allow for +<br/>evaluation of system issues including: +<br/> Long distance recognition +<br/> Blur caused by atmospherics +<br/> Various weather conditions +<br/> End to end systems evaluation +<br/>Our contributions: +<br/> Define a taxonomy of face-models +<br/>for controlled experimentations +<br/> Show how Synthetic addresses gaps +<br/>in system evaluation +<br/> Show a process for generating and +<br/>validating synthetic models +<br/> Use these models in long distance +<br/>face recognition system evaluation +<br/>Experimental Setup +<br/>Results and Conclusions +<br/>Example Models +<br/>Original Pie +<br/>Semi- +<br/>Synthetic +<br/>FaceGen +<br/>Animetrics +<br/>http://www.facegen.com +<br/>http://www.animetrics.com/products/Forensica.php +<br/>Guided- +<br/>Synthetic +<br/>Models +<br/> Models generated using the well +<br/>known CMU PIE [18] dataset. Each of +<br/>the 68 subjects of PIE were modeled +<br/>using a right profile and frontal +<br/>image from the lights subset. +<br/> Two modeling programs were used, +<br/>Facegen and Animetrics. Both +<br/>programs create OBJ files and +<br/>textures +<br/> Models are re-rendered using +<br/>custom display software built with +<br/>OpenGL, GLUT and DevIL libraries +<br/> Custom Display Box housing a BENQ SP820 high +<br/>powered projector rated at 4000 ANSI Lumens +<br/> Canon EOS 7D withd a Sigma 800mm F5.6 EX APO +<br/>DG HSM lens a 2x adapter imaging the display +<br/>from 214 meters +<br/>Normalized Example Captures +<br/>Real PIE 1 Animetrics +<br/>FaceGen +<br/>81M inside 214M outside +<br/>Real PIE 2 +<br/> Pre-cropped images were used for the +<br/>commercial core +<br/> Ground truth eye points + geometric/lighting +<br/>normalization pre processing before running +<br/>through the implementation of the V1 +<br/>recognition algorithm found in [1]. +<br/> Geo normalization highlights how the feature +<br/>region of the models looks very similar to +<br/>that of the real person. +<br/>Each test consisted of using 3 approximately frontal gallery images NOT used to +<br/>make the 3D model used as the probe, best score over 3 images determined score. +<br/>Even though the PIE-3D-20100224A–D sets were imaged on the same day, the V1 +<br/>core scored differently on each highlighting the synthetic data’s ability to help +<br/>evaluate data capture methods and effects of varying atmospherics. The ISO setting +<br/>varied which effects the shutter speed, with higher ISO generally yielding less blur. +<br/>Dataset +<br/>Range(m) +<br/>Iso +<br/>V1 +<br/>Comm. +<br/>Original PIE Images +<br/>FaceGen ScreenShots +<br/>Animetrics Screenshots +<br/>PIE-3D-20100210B +<br/>PIE-3D-20100224A +<br/>PIE-3D-20100224B +<br/>PIE-3D-20100224C +<br/>PIE-3D-20100224D +<br/>N/A +<br/>N/A +<br/>N/A +<br/>81m +<br/>214m +<br/>214m +<br/>214m +<br/>214m +<br/>N/A +<br/>N/A +<br/>N/A +<br/>500 +<br/>125 +<br/>125 +<br/>250 +<br/>400 +<br/>100 +<br/>47.76 +<br/>100 +<br/>100 +<br/>58.82 +<br/>45.59 +<br/>81.82 +<br/>79.1 +<br/>100 +<br/>100 +<br/>100 +<br/>100 +<br/>100 +<br/>100 +<br/> The same (100 percent) recognition rate on screenshots as original images +<br/>validate the Anmetrics guided synthetic models and fails FaceGen Models. +<br/> 100% recognition means dataset is too small/easy; exapanding pose and models +<br/>underway. +<br/> Expanded the photohead methodology into 3D +<br/> Developed a robust modeling system allowing for multiple configurations of a +<br/>single real life data set. +<br/> Gabor+SVM based V1[15] significantly more impacted by atmospheric blur than +<br/>the commercial algorithm +<br/>Key References: +<br/>[6 of 21] R. Bevridge, D. Bolme, M Teixeira, and B. Draper. The CSU Face Identification Evaluation System Users Guide: Version 5.0. Technical report, CSU 2003 +<br/>[8 of 21] T. Boult and W. Scheirer. Long range facial image acquisition and quality. In M. Tisarelli, S. Li, and R. Chellappa. +<br/>[15 of 21] N. Pinto, J. J. DiCarlo, and D. D. Cox. How far can you get with a modern face recognition test set using only simple features? In IEEE CVPR, 2009. +<br/>[18 of 21] T. Sim, S. Baker, and M. Bsat. The CMU Pose, Illumination and Expression (PIE) Database. In Proceedings of the IEEE F&G, May 2002. +</td></tr><tr><td>5721216f2163d026e90d7cd9942aeb4bebc92334</td><td></td></tr><tr><td>5753b2b5e442eaa3be066daa4a2ca8d8a0bb1725</td><td></td></tr><tr><td>574ad7ef015995efb7338829a021776bf9daaa08</td><td>AdaScan: Adaptive Scan Pooling in Deep Convolutional Neural Networks +<br/>for Human Action Recognition in Videos +<br/>1IIT Kanpur‡ +<br/>2SRI International +<br/>3UCSD +</td></tr><tr><td>57d37ad025b5796457eee7392d2038910988655a</td><td>GEERATVEEETATF +<br/> +<br/>by +<br/>DagaEha +<br/>UdeheS eviif +<br/>f.DahaWeiha +<br/>ATheiS biediaia +<br/>Re ieefheDegeef +<br/>aefSciece +<br/>a +<br/>TheSch +<br/> +<br/>Decebe2009 +</td></tr><tr><td>3b1260d78885e872cf2223f2c6f3d6f6ea254204</td><td></td></tr><tr><td>3b1aaac41fc7847dd8a6a66d29d8881f75c91ad5</td><td>Sparse Representation-based Open Set Recognition +</td></tr><tr><td>3bc776eb1f4e2776f98189e17f0d5a78bb755ef4</td><td></td></tr><tr><td>3b4fd2aec3e721742f11d1ed4fa3f0a86d988a10</td><td>Glimpse: Continuous, Real-Time Object Recognition on +<br/>Mobile Devices +<br/>MIT CSAIL +<br/>Microsoft Research +<br/>MIT CSAIL +<br/>Microsoft Research +<br/>MIT CSAIL +</td></tr><tr><td>3b15a48ffe3c6b3f2518a7c395280a11a5f58ab0</td><td>On Knowledge Transfer in +<br/>Object Class Recognition +<br/>A dissertation approved by +<br/>TECHNISCHE UNIVERSITÄT DARMSTADT +<br/>Fachbereich Informatik +<br/>for the degree of +<br/>Doktor-Ingenieur (Dr.-Ing.) +<br/>presented by +<br/>Dipl.-Inform. +<br/>born in Mainz, Germany +<br/>Prof. Dr.-Ing. Michael Goesele, examiner +<br/>Prof. Martial Hebert, Ph.D., co-examiner +<br/>Prof. Dr. Bernt Schiele, co-examiner +<br/>Date of Submission: 12th of August, 2010 +<br/>Date of Defense: 23rd of September, 2010 +<br/>Darmstadt, 2010 +<br/>D17 +</td></tr><tr><td>3ba8f8b6bfb36465018430ffaef10d2caf3cfa7e</td><td>Local Directional Number Pattern for Face +<br/>Analysis: Face and Expression Recognition +</td></tr><tr><td>3b80bf5a69a1b0089192d73fa3ace2fbb52a4ad5</td><td></td></tr><tr><td>3b9d94752f8488106b2c007e11c193f35d941e92</td><td>CVPR +<br/>#2052 +<br/>000 +<br/>001 +<br/>002 +<br/>003 +<br/>004 +<br/>005 +<br/>006 +<br/>007 +<br/>008 +<br/>009 +<br/>010 +<br/>011 +<br/>012 +<br/>013 +<br/>014 +<br/>015 +<br/>016 +<br/>017 +<br/>018 +<br/>019 +<br/>020 +<br/>021 +<br/>022 +<br/>023 +<br/>024 +<br/>025 +<br/>026 +<br/>027 +<br/>028 +<br/>029 +<br/>030 +<br/>031 +<br/>032 +<br/>033 +<br/>034 +<br/>035 +<br/>036 +<br/>037 +<br/>038 +<br/>039 +<br/>040 +<br/>041 +<br/>042 +<br/>043 +<br/>044 +<br/>045 +<br/>046 +<br/>047 +<br/>048 +<br/>049 +<br/>050 +<br/>051 +<br/>052 +<br/>053 +<br/>CVPR 2013 Submission #2052. CONFIDENTIAL REVIEW COPY. DO NOT DISTRIBUTE. +<br/>CVPR +<br/>#2052 +<br/>Appearance, Visual and Social Ensembles for +<br/>Face Recognition in Personal Photo Collections +<br/>Anonymous CVPR submission +<br/>Paper ID 2052 +</td></tr><tr><td>3be7b7eb11714e6191dd301a696c734e8d07435f</td><td></td></tr><tr><td>3b410ae97e4564bc19d6c37bc44ada2dcd608552</td><td>Scalability Analysis of Audio-Visual Person +<br/>Identity Verification +<br/>1 Communications Laboratory, +<br/>Universit´e catholique de Louvain, B-1348 Belgium, +<br/>2 IDIAP, CH-1920 Martigny, +<br/>Switzerland +</td></tr><tr><td>6f5ce5570dc2960b8b0e4a0a50eab84b7f6af5cb</td><td>Low Resolution Face Recognition Using a +<br/>Two-Branch Deep Convolutional Neural Network +<br/>Architecture +</td></tr><tr><td>6f288a12033fa895fb0e9ec3219f3115904f24de</td><td>Learning Expressionlets via Universal Manifold +<br/>Model for Dynamic Facial Expression Recognition +</td></tr><tr><td>6f2dc51d607f491dbe6338711c073620c85351ac</td><td></td></tr><tr><td>6f75697a86d23d12a14be5466a41e5a7ffb79fad</td><td></td></tr><tr><td>6f7d06ced04ead3b9a5da86b37e7c27bfcedbbdd</td><td>Pages 51.1-51.12 +<br/>DOI: https://dx.doi.org/10.5244/C.30.51 +</td></tr><tr><td>6f7a8b3e8f212d80f0fb18860b2495be4c363eac</td><td>Creating Capsule Wardrobes from Fashion Images +<br/>UT-Austin +<br/>UT-Austin +</td></tr><tr><td>6f6b4e2885ea1d9bea1bb2ed388b099a5a6d9b81</td><td>Structured Output SVM Prediction of Apparent Age, +<br/>Gender and Smile From Deep Features +<br/>Michal Uˇriˇc´aˇr +<br/>CMP, Dept. of Cybernetics +<br/>FEE, CTU in Prague +<br/>Computer Vision Lab +<br/>D-ITET, ETH Zurich +<br/>Computer Vision Lab +<br/>D-ITET, ETH Zurich +<br/>PSI, ESAT, KU Leuven +<br/>CVL, D-ITET, ETH Zurich +<br/>Jiˇr´ı Matas +<br/>CMP, Dept. of Cybernetics +<br/>FEE, CTU in Prague +</td></tr><tr><td>6f35b6e2fa54a3e7aaff8eaf37019244a2d39ed3</td><td>DOI 10.1007/s00530-005-0177-4 +<br/>R E G U L A R PA P E R +<br/>Learning probabilistic classifiers for human–computer +<br/>interaction applications +<br/>Published online: 10 May 2005 +<br/>c(cid:1) Springer-Verlag 2005 +<br/>intelligent +<br/>interaction, +</td></tr><tr><td>6fa3857faba887ed048a9e355b3b8642c6aab1d8</td><td>Face Recognition in Challenging Environments: +<br/>An Experimental and Reproducible Research +<br/>Survey +</td></tr><tr><td>6f7ce89aa3e01045fcd7f1c1635af7a09811a1fe</td><td>978-1-4673-0046-9/12/$26.00 ©2012 IEEE +<br/>937 +<br/>ICASSP 2012 +</td></tr><tr><td>6fe2efbcb860767f6bb271edbb48640adbd806c3</td><td>SOFT BIOMETRICS: HUMAN IDENTIFICATION USING COMPARATIVE DESCRIPTIONS +<br/>Soft Biometrics; Human Identification using +<br/>Comparative Descriptions +</td></tr><tr><td>6fdc0bc13f2517061eaa1364dcf853f36e1ea5ae</td><td>DAISEE: Dataset for Affective States in +<br/>E-Learning Environments +<br/>1 Microsoft India R&D Pvt. Ltd. +<br/>2 Department of Computer Science, IIT Hyderabad +</td></tr><tr><td>6f5151c7446552fd6a611bf6263f14e729805ec7</td><td>5KHHAO /7 %:0 7 +<br/>)>IJH=?J 9EJDE JDA ?JANJ B=?A ANFHAIIE ?=IIE?=JE KIEC JDA +<br/>FH>=>EEJEAI JD=J A=?D A B IALAH= ?O ??KHHEC )7 CHKFI EI +<br/>?=IIIAF=H=>EEJO MAECDJEC +<br/>/=>H M=LAAJI H FHE?EF= ?FAJI ==OIEI 2+) ! 1 JDEI F=FAH MA +</td></tr><tr><td>03c56c176ec6377dddb6a96c7b2e95408db65a7a</td><td>A Novel Geometric Framework on Gram Matrix +<br/>Trajectories for Human Behavior Understanding +</td></tr><tr><td>03d9ccce3e1b4d42d234dba1856a9e1b28977640</td><td></td></tr><tr><td>0322e69172f54b95ae6a90eb3af91d3daa5e36ea</td><td>Face Classification using Adjusted Histogram in +<br/>Grayscale +</td></tr><tr><td>03f7041515d8a6dcb9170763d4f6debd50202c2b</td><td>Clustering Millions of Faces by Identity +</td></tr><tr><td>038ce930a02d38fb30d15aac654ec95640fe5cb0</td><td>Approximate Structured Output Learning for Constrained Local +<br/>Models with Application to Real-time Facial Feature Detection and +<br/>Tracking on Low-power Devices +</td></tr><tr><td>03c1fc9c3339813ed81ad0de540132f9f695a0f8</td><td>Proceedings of Machine Learning Research 81:1–15, 2018 +<br/>Conference on Fairness, Accountability, and Transparency +<br/>Gender Shades: Intersectional Accuracy Disparities in +<br/>Commercial Gender Classification∗ +<br/>MIT Media Lab 75 Amherst St. Cambridge, MA 02139 +<br/>Microsoft Research 641 Avenue of the Americas, New York, NY 10011 +<br/>Editors: Sorelle A. Friedler and Christo Wilson +</td></tr><tr><td>0339459a5b5439d38acd9c40a0c5fea178ba52fb</td><td>D|C|I&I 2009 Prague +<br/>Multimodal recognition of emotions in car +<br/>environments +</td></tr><tr><td>03a8f53058127798bc2bc0245d21e78354f6c93b</td><td>Max-Margin Additive Classifiers for Detection +<br/>Sam Hare +<br/>VGG Reading Group +<br/>October 30, 2009 +</td></tr><tr><td>03fc466fdbc8a2efb6e3046fcc80e7cb7e86dc20</td><td>A Real Time System for Model-based Interpretation of +<br/>the Dynamics of Facial Expressions +<br/>Technische Universit¨at M¨unchen +<br/>Boltzmannstr. 3, 85748 Garching +<br/>1. Motivation +<br/>Recent progress in the field of Computer Vision allows +<br/>intuitive interaction via speech, gesture or facial expressions +<br/>between humans and technical systems.Model-based tech- +<br/>niques facilitate accurately interpreting images with faces +<br/>by exploiting a priori knowledge, such as shape and texture +<br/>information. This renders them an inevitable component +<br/>to realize the paradigm of intuitive human-machine interac- +<br/>tion. +<br/>Our demonstration shows model-based recognition of +<br/>facial expressions in real-time via the state-of-the-art +<br/>Candide-3 face model [1] as visible in Figure 1. This three- +<br/>dimensional and deformable model is highly appropriate +<br/>for real-world face interpretation applications. However, +<br/>its complexity challenges the task of model fitting and we +<br/>tackle this challenge with an algorithm that has been auto- +<br/>matically learned from a large set of images. This solution +<br/>provides both, high accuracy and runtime. Note, that our +<br/>system is not limited to facial expression estimation. Gaze +<br/>direction, gender and age are also estimated. +<br/>2. Face Model Fitting +<br/>Models reduce the large amount of image data to a +<br/>small number of model parameters to describe the im- +<br/>age content, which facilitates and accelerates the subse- +<br/>quent interpretation task. Cootes et al. [3] introduced mod- +<br/>elling shapes with Active Contours. Further enhancements +<br/>emerged the idea of expanding shape models with texture +<br/>information [2]. Recent research considers modelling faces +<br/>in 3D space [1, 10]. +<br/>Fitting the face model is the computational challenge of +<br/>finding the parameters that best describe the face within a +<br/>given image. This task is often addressed by minimizing +<br/>an objective function, such as the pixel error between the +<br/>model’s rendered surface and the underlying image content. +<br/>This section describes the four main components of model- +<br/>based techniques, see [9]. +<br/>The face model contains a parameter vector p that repre- +<br/>sents its configurations. We integrate the complex and de- +<br/>formable 3D wire frame Candide-3 face model [1]. The +<br/>model consists of 116 anatomical landmarks and its param- +<br/>eter vector p = (rx, ry, rz, s, tx, ty, σ, α)T describes the +<br/>affine transformation (rx, ry, rz, s, tx, ty) and the deforma- +<br/>tion (σ, α). The 79 deformation parameters indicate the +<br/>shape of facial components such as the mouth, the eyes, or +<br/>the eye brows, etc., see Figure 2. +<br/>The localization algorithm computes an initial estimate of +<br/>the model parameters that is further refined by the subse- +<br/>quent fitting algorithm. Our system integrates the approach +<br/>of [8], which detects the model’s affine transformation in +<br/>case the image shows a frontal view face. +<br/>The objective function yields a comparable value that +<br/>specifies how accurately a parameterized model matches an +<br/>image. Traditional approaches manually specify the objec- +<br/>tive function in a laborious and erroneous task. In contrast, +<br/>we automatically learn the objective function from a large +<br/>set of training data based on objective information theoretic +<br/>measures [9]. This approach does not require expert knowl- +<br/>edge and it is domain-independently applicable. As a re- +<br/>sult, this approach yields more robust and accurate objective +<br/>functions, which greatly facilitate the task of the associated +<br/>fitting algorithms. Accurately estimated model parameters +<br/>in turn are required to infer correct high-level information, +<br/>such as facial expression or gaze direction. +<br/>Figure 1. Interpreting expressions with the Candide-3 face model. +</td></tr><tr><td>03b98b4a2c0b7cc7dae7724b5fe623a43eaf877b</td><td>Acume: A Novel Visualization Tool for Understanding Facial +<br/>Expression and Gesture Data +</td></tr><tr><td>03104f9e0586e43611f648af1132064cadc5cc07</td><td></td></tr><tr><td>03f14159718cb495ca50786f278f8518c0d8c8c9</td><td>2015 IEEE International Conference on Control System, Computing and Engineering, Nov 27 – Nov 29, 2015 Penang, Malaysia +<br/>2015 IEEE International Conference on Control System, +<br/>Computing and Engineering (ICCSCE2015) +<br/>Technical Session 1A – DAY 1 – 27th Nov 2015 +<br/>Time: 3.00 pm – 4.30 pm +<br/>Venue: Jintan +<br/>Topic: Signal and Image Processing +<br/>3.00 pm – 3.15pm +<br/>3.15 pm – 3.30pm +<br/>3.30 pm – 3.45pm +<br/>3.45 pm – 4.00pm +<br/>4.00 pm – 4.15pm +<br/>4.15 pm – 4.30pm +<br/>4.30 pm – 4.45pm +<br/>1A 01 ID3 +<br/>Can Subspace Based Learning Approach Perform on Makeup Face +<br/>Recognition? +<br/>Khor Ean Yee, Pang Ying Han, Ooi Shih Yin and Wee Kuok Kwee +<br/>1A 02 ID35 +<br/>Performance Evaluation of HOG and Gabor Features for Vision-based +<br/>Vehicle Detection +<br/>1A 03 ID23 +<br/>Experimental Method to Pre-Process Fuzzy Bit Planes before Low-Level +<br/>Feature Extraction in Thermal Images +<br/>Chan Wai Ti and Sim Kok Swee +<br/>1A 04 ID84 +<br/>Fractal-based Texture and HSV Color Features for Fabric Image Retrieval +<br/>Nanik Suciati, Darlis Herumurti and Arya Yudhi Wijaya +<br/>1A 05 ID168 +<br/>Study of Automatic Melody Extraction Methods for Philippine Indigenous +<br/>Music +<br/>Jason Disuanco, Vanessa Tan, Franz de Leon +<br/>1A 06 ID211 +<br/>Acoustical Comparison between Voiced and Voiceless Arabic Phonemes of +<br/>Malay +<br/>Speakers +<br/>Ali Abd Almisreb, Ahmad Farid Abidin, Nooritawati Md Tahir +<br/>*shaded cell is the proposed session chair +<br/>viii +<br/>©Faculty of Electrical Engineering, Universiti Teknologi MARA +</td></tr><tr><td>0394040749195937e535af4dda134206aa830258</td><td>Geodesic Entropic Graphs for Dimension and +<br/>Entropy Estimation in Manifold Learning +<br/>December 16, 2003 +</td></tr><tr><td>0334cc0374d9ead3dc69db4816d08c917316c6c4</td><td></td></tr><tr><td>0394e684bd0a94fc2ff09d2baef8059c2652ffb0</td><td>Median Robust Extended Local Binary Pattern +<br/>for Texture Classification +<br/>Index Terms— Texture descriptors, rotation invariance, local +<br/>binary pattern (LBP), feature extraction, texture analysis. +<br/>how the texture recognition process works in humans as +<br/>well as in the important role it plays in the wide variety of +<br/>applications of computer vision and image analysis [1], [2]. +<br/>The many applications of texture classification include medical +<br/>image analysis and understanding, object recognition, biomet- +<br/>rics, content-based image retrieval, remote sensing, industrial +<br/>inspection, and document classification. +<br/>As a classical pattern recognition problem, texture classifi- +<br/>cation primarily consists of two critical subproblems: feature +<br/>extraction and classifier designation [1], [2]. It is generally +<br/>agreed that the extraction of powerful texture features plays a +<br/>relatively more important role, since if poor features are used +<br/>even the best classifier will fail to achieve good recognition +<br/>results. Consequently, most research in texture classification +<br/>focuses on the feature extraction part and numerous texture +<br/>feature extraction methods have been developed, with excellent +<br/>surveys given in [1]–[5]. Most existing methods have not, +<br/>however, been capable of performing sufficiently well for +<br/>real-world applications, which have demanding requirements +<br/>including database size, nonideal environmental conditions, +<br/>and running in real-time. +</td></tr><tr><td>03e88bf3c5ddd44ebf0e580d4bd63072566613ad</td><td></td></tr><tr><td>03f4c0fe190e5e451d51310bca61c704b39dcac8</td><td>J Ambient Intell Human Comput +<br/>DOI 10.1007/s12652-016-0406-z +<br/>O R I G I N A L R E S E A R C H +<br/>CHEAVD: a Chinese natural emotional audio–visual database +<br/>Received: 30 March 2016 / Accepted: 22 August 2016 +<br/>Ó Springer-Verlag Berlin Heidelberg 2016 +</td></tr><tr><td>031055c241b92d66b6984643eb9e05fd605f24e2</td><td>Multi-fold MIL Training for Weakly Supervised Object Localization +<br/>Inria∗ +</td></tr><tr><td>0332ae32aeaf8fdd8cae59a608dc8ea14c6e3136</td><td>Int J Comput Vis +<br/>DOI 10.1007/s11263-017-1009-7 +<br/>Large Scale 3D Morphable Models +<br/>Received: 15 March 2016 / Accepted: 24 March 2017 +<br/>© The Author(s) 2017. This article is an open access publication +</td></tr><tr><td>034addac4637121e953511301ef3a3226a9e75fd</td><td>Implied Feedback: Learning Nuances of User Behavior in Image Search +<br/>Virginia Tech +</td></tr><tr><td>03701e66eda54d5ab1dc36a3a6d165389be0ce79</td><td>179 +<br/>Improved Principal Component Regression for Face +<br/>Recognition Under Illumination Variations +</td></tr><tr><td>9b318098f3660b453fbdb7a579778ab5e9118c4c</td><td>3931 +<br/>Joint Patch and Multi-label Learning for Facial +<br/>Action Unit and Holistic Expression Recognition +<br/>classifiers without +</td></tr><tr><td>9b000ccc04a2605f6aab867097ebf7001a52b459</td><td></td></tr><tr><td>9b474d6e81e3b94e0c7881210e249689139b3e04</td><td>VG-RAM Weightless Neural Networks for +<br/>Face Recognition +<br/>Departamento de Inform´atica +<br/>Universidade Federal do Esp´ırito Santo +<br/>Av. Fernando Ferrari, 514, 29075-910 - Vit´oria-ES +<br/>Brazil +<br/>1. Introduction +<br/>Computerized human face recognition has many practical applications, such as access control, +<br/>security monitoring, and surveillance systems, and has been one of the most challenging and +<br/>active research areas in computer vision for many decades (Zhao et al.; 2003). Even though +<br/>current machine recognition systems have reached a certain level of maturity, the recognition +<br/>of faces with different facial expressions, occlusions, and changes in illumination and/or pose +<br/>is still a hard problem. +<br/>A general statement of the problem of machine recognition of faces can be formulated as fol- +<br/>lows: given an image of a scene, (i) identify or (ii) verify one or more persons in the scene +<br/>using a database of faces. In identification problems, given a face as input, the system reports +<br/>back the identity of an individual based on a database of known individuals; whereas in veri- +<br/>fication problems, the system confirms or rejects the claimed identity of the input face. In both +<br/>cases, the solution typically involves segmentation of faces from scenes (face detection), fea- +<br/>ture extraction from the face regions, recognition, or verification. In this chapter, we examine +<br/>the recognition of frontal face images required in the context of identification problems. +<br/>Many approaches have been proposed to tackle the problem of face recognition. One can +<br/>roughly divide these into (i) holistic approaches, (ii) feature-based approaches, and (iii) hybrid +<br/>approaches (Zhao et al.; 2003). Holistic approaches use the whole face region as the raw input +<br/>to a recognition system (a classifier). In feature-based approaches, local features, such as the +<br/>eyes, nose, and mouth, are first extracted and their locations and local statistics (geometric +<br/>and/or appearance based) are fed into a classifier. Hybrid approaches use both local features +<br/>and the whole face region to recognize a face. +<br/>Among +<br/>fisher- +<br/>faces (Belhumeur et al.; 1997; Etemad and Chellappa; 1997) have proved to be effective +<br/>(Turk and Pentland; +<br/>eigenfaces +<br/>holistic +<br/>approaches, +<br/>1991) +<br/>and +</td></tr><tr><td>9bc01fa9400c231e41e6a72ec509d76ca797207c</td><td></td></tr><tr><td>9bcfadd22b2c84a717c56a2725971b6d49d3a804</td><td>How to Detect a Loss of Attention in a Tutoring System +<br/>using Facial Expressions and Gaze Direction +</td></tr><tr><td>9bac481dc4171aa2d847feac546c9f7299cc5aa0</td><td>Matrix Product State for Higher-Order Tensor +<br/>Compression and Classification +</td></tr><tr><td>9b7974d9ad19bb4ba1ea147c55e629ad7927c5d7</td><td>Faical Expression Recognition by Combining +<br/>Texture and Geometrical Features +</td></tr><tr><td>9ea73660fccc4da51c7bc6eb6eedabcce7b5cead</td><td>Talking Head Detection by Likelihood-Ratio Test† +<br/>MIT Lincoln Laboratory, +<br/>Lexington MA 02420, USA +</td></tr><tr><td>9e9052256442f4e254663ea55c87303c85310df9</td><td>International Journal of Advanced Research in Computer Engineering & Technology (IJARCET) +<br/>Volume 4 Issue 10, October 2015 +<br/>Review On Attribute-assisted Reranking for +<br/>Image Search +<br/> +</td></tr><tr><td>9e0285debd4b0ba7769b389181bd3e0fd7a02af6</td><td>From face images and attributes to attributes +<br/>Computer Vision Laboratory, ETH Zurich, Switzerland +</td></tr><tr><td>9e5c2d85a1caed701b68ddf6f239f3ff941bb707</td><td></td></tr><tr><td>04bb3fa0824d255b01e9db4946ead9f856cc0b59</td><td></td></tr><tr><td>040dc119d5ca9ea3d5fc39953a91ec507ed8cc5d</td><td>Noname manuscript No. +<br/>(will be inserted by the editor) +<br/>Large-scale Bisample Learning on ID vs. Spot Face Recognition +<br/>Received: date / Accepted: date +</td></tr><tr><td>04470861408d14cc860f24e73d93b3bb476492d0</td><td></td></tr><tr><td>0447bdb71490c24dd9c865e187824dee5813a676</td><td>Manifold Estimation in View-based Feature +<br/>Space for Face Synthesis Across Pose +<br/>Paper 27 +</td></tr><tr><td>044ba70e6744e80c6a09fa63ed6822ae241386f2</td><td>TO APPEAR IN AUTONOMOUS ROBOTS, SPECIAL ISSUE IN LEARNING FOR HUMAN-ROBOT COLLABORATION +<br/>Early Prediction for Physical Human Robot +<br/>Collaboration in the Operating Room +</td></tr><tr><td>04dcdb7cb0d3c462bdefdd05508edfcff5a6d315</td><td>Assisting the training of deep neural networks +<br/>with applications to computer vision +<br/>tesi doctoral està subjecta a +<br/>la +<br/>Aquesta +<br/>CompartirIgual 4.0. Espanya de Creative Commons. +<br/>Esta tesis doctoral está sujeta a la licencia Reconocimiento - NoComercial – CompartirIgual +<br/>4.0. España de Creative Commons. +<br/>This doctoral thesis is licensed under the Creative Commons Attribution-NonCommercial- +<br/>ShareAlike 4.0. Spain License. +<br/>llicència Reconeixement- NoComercial – +</td></tr><tr><td>044fdb693a8d96a61a9b2622dd1737ce8e5ff4fa</td><td>Dynamic Texture Recognition Using Local Binary +<br/>Patterns with an Application to Facial Expressions +</td></tr><tr><td>04250e037dce3a438d8f49a4400566457190f4e2</td><td></td></tr><tr><td>0431e8a01bae556c0d8b2b431e334f7395dd803a</td><td>Learning Localized Perceptual Similarity Metrics for Interactive Categorization +<br/>Google Inc. +<br/>google.com +</td></tr><tr><td>04b4c779b43b830220bf938223f685d1057368e9</td><td>Video retrieval based on deep convolutional +<br/>neural network +<br/>Yajiao Dong +<br/>School of Information and Electronics, +<br/>Beijing Institution of Technology, Beijing, China +<br/>Jianguo Li +<br/>School of Information and Electronics, +<br/>Beijing Institution of Technology, Beijing, China +</td></tr><tr><td>04616814f1aabe3799f8ab67101fbaf9fd115ae4</td><td><b>UNIVERSIT´EDECAENBASSENORMANDIEU.F.R.deSciences´ECOLEDOCTORALESIMEMTH`ESEPr´esent´eeparM.GauravSHARMAsoutenuele17D´ecembre2012envuedel’obtentionduDOCTORATdel’UNIVERSIT´EdeCAENSp´ecialit´e:InformatiqueetapplicationsArrˆet´edu07aoˆut2006Titre:DescriptionS´emantiquedesHumainsPr´esentsdansdesImagesVid´eo(SemanticDescriptionofHumansinImages)TheworkpresentedinthisthesiswascarriedoutatGREYC-UniversityofCaenandLEAR–INRIAGrenobleJuryM.PatrickPEREZDirecteurdeRechercheINRIA/Technicolor,RennesRapporteurM.FlorentPERRONNINPrincipalScientistXeroxRCE,GrenobleRapporteurM.JeanPONCEProfesseurdesUniversit´esENS,ParisExaminateurMme.CordeliaSCHMIDDirectricedeRechercheINRIA,GrenobleDirectricedeth`eseM.Fr´ed´ericJURIEProfesseurdesUniversit´esUniversit´edeCaenDirecteurdeth`ese</b></td></tr><tr><td>6a3a07deadcaaab42a0689fbe5879b5dfc3ede52</td><td>Learning to Estimate Pose by Watching Videos +<br/>Department of Computer Science and Engineering +<br/>IIT Kanpur +</td></tr><tr><td>6ad107c08ac018bfc6ab31ec92c8a4b234f67d49</td><td></td></tr><tr><td>6a184f111d26787703f05ce1507eef5705fdda83</td><td></td></tr><tr><td>6a16b91b2db0a3164f62bfd956530a4206b23fea</td><td>A Method for Real-Time Eye Blink Detection and Its Application +<br/>Mahidol Wittayanusorn School +<br/>Puttamonton, Nakornpatom 73170, Thailand +</td></tr><tr><td>6a806978ca5cd593d0ccd8b3711b6ef2a163d810</td><td>Facial feature tracking for Emotional Dynamic +<br/>Analysis +<br/>1ISIR, CNRS UMR 7222 +<br/>Univ. Pierre et Marie Curie, Paris +<br/>2LAMIA, EA 4540 +<br/>Univ. of Fr. West Indies & Guyana +</td></tr><tr><td>6a8a3c604591e7dd4346611c14dbef0c8ce9ba54</td><td>ENTERFACE’10, JULY 12TH - AUGUST 6TH, AMSTERDAM, THE NETHERLANDS. +<br/>58 +<br/>An Affect-Responsive Interactive Photo Frame +</td></tr><tr><td>6aa43f673cc42ed2fa351cbc188408b724cb8d50</td><td></td></tr><tr><td>6aefe7460e1540438ffa63f7757c4750c844764d</td><td>Non-rigid Segmentation using Sparse Low Dimensional Manifolds and +<br/>Deep Belief Networks ∗ +<br/>Instituto de Sistemas e Rob´otica +<br/>Instituto Superior T´ecnico, Portugal +</td></tr><tr><td>6a1beb34a2dfcdf36ae3c16811f1aef6e64abff2</td><td></td></tr><tr><td>322c063e97cd26f75191ae908f09a41c534eba90</td><td>Noname manuscript No. +<br/>(will be inserted by the editor) +<br/>Improving Image Classification using Semantic Attributes +<br/>Received: date / Accepted: date +</td></tr><tr><td>325b048ecd5b4d14dce32f92bff093cd744aa7f8</td><td>CVPR +<br/>#2670 +<br/>000 +<br/>001 +<br/>002 +<br/>003 +<br/>004 +<br/>005 +<br/>006 +<br/>007 +<br/>008 +<br/>009 +<br/>010 +<br/>011 +<br/>012 +<br/>013 +<br/>014 +<br/>015 +<br/>016 +<br/>017 +<br/>018 +<br/>019 +<br/>020 +<br/>021 +<br/>022 +<br/>023 +<br/>024 +<br/>025 +<br/>026 +<br/>027 +<br/>028 +<br/>029 +<br/>030 +<br/>031 +<br/>032 +<br/>033 +<br/>034 +<br/>035 +<br/>036 +<br/>037 +<br/>038 +<br/>039 +<br/>040 +<br/>041 +<br/>042 +<br/>043 +<br/>044 +<br/>045 +<br/>046 +<br/>047 +<br/>048 +<br/>049 +<br/>050 +<br/>051 +<br/>052 +<br/>053 +<br/>CVPR 2008 Submission #2670. CONFIDENTIAL REVIEW COPY. DO NOT DISTRIBUTE. +<br/>CVPR +<br/>#2670 +<br/>Multi-Image Graph Cut Clothing Segmentation for Recognizing People +<br/>Anonymous CVPR submission +<br/>Paper ID 2670 +</td></tr><tr><td>321bd4d5d80abb1bae675a48583f872af3919172</td><td>Wang et al. EURASIP Journal on Image and Video Processing (2016) 2016:44 +<br/>DOI 10.1186/s13640-016-0152-3 +<br/>EURASIP Journal on Image +<br/>and Video Processing +<br/>R EV I E W +<br/>Entropy-weighted feature-fusion method +<br/>for head-pose estimation +<br/>Open Access +</td></tr><tr><td>32b8c9fd4e3f44c371960eb0074b42515f318ee7</td><td></td></tr><tr><td>32575ffa69d85bbc6aef5b21d73e809b37bf376d</td><td>-)5741/ *1-641+ 5)2- 37)16; 1 6-45 . *1-641+ 1.4)61 +<br/>7ELAHIEJO B JJ=M= +<br/>)*564)+6 +<br/>IKHA L=HE=JEI E >EAJHE? I=FA GK=EJO 9A >ACE MEJD +<br/>IKHAAJI 9A JDA IDM JD=J JDA >EAJHE? EBH=JE BH +<br/>JA EI JDA A= D(p(cid:107)q) BH = FAHII E JDA FFK=JE 1 +<br/>BH I= ALAI B >KH MEJD = =IOFJJE? >AD=LEH =J =HCAH +<br/>>KH +<br/> 164,7+61 +<br/>*EAJHE? I=FA GK=EJO EI = A=IKHA B JDA KIABKAII B = +<br/>GK=EJO +<br/>F=FAH MA FHFIA = AM =FFH=?D J A=IKHA JDEI GK=JEJO +<br/>JDA EJKEJELA >IAHL=JE JD=J = DECD GK=EJO >EAJHE? E=CA +<br/>>EAJHE? EBH=JE +<br/>EIIKAI E >EAJHE? JA?DCO .H AN=FA A B JDA IJ +<br/>? >EAJHE? GKAIJEI EI JD=J B KEGKAAII AC J MD=J +<br/>ANJAJ =HA CAHFHEJI KEGKA .H JDA FEJ B LEAM B +<br/>=>A EBH=JE EI =L=E=>A BH = CELA JA?DCO IK?D +<br/> $ " +<br/>1 JDEI F=FAH MA A=>H=JA = =FFH=?D J +<br/>BMI +<br/>AJI +<br/> >ABHA = >EAJHE? A=IKHAAJ t0 =J MDE?D JEA MA O +<br/>M = FAHI p EI F=HJ B = FFK=JE q MDE?D =O >A JDA +</td></tr><tr><td>324b9369a1457213ec7a5a12fe77c0ee9aef1ad4</td><td>Dynamic Facial Analysis: From Bayesian Filtering to Recurrent Neural Network +<br/>NVIDIA +</td></tr><tr><td>32df63d395b5462a8a4a3c3574ae7916b0cd4d1d</td><td>978-1-4577-0539-7/11/$26.00 ©2011 IEEE +<br/>1489 +<br/>ICASSP 2011 +</td></tr><tr><td>35308a3fd49d4f33bdbd35fefee39e39fe6b30b7</td><td></td></tr><tr><td>352d61eb66b053ae5689bd194840fd5d33f0e9c0</td><td>Analysis Dictionary Learning based +<br/>Classification: Structure for Robustness +</td></tr><tr><td>3538d2b5f7ab393387ce138611ffa325b6400774</td><td>A DSP-BASED APPROACH FOR THE IMPLEMENTATION OF FACE RECOGNITION +<br/>ALGORITHMS +<br/>A. U. Batur +<br/>B. E. Flinchbaugh +<br/>M. H. Hayes IIl +<br/>Center for Signal and Image Proc. +<br/>Georgia Inst. Of Technology +<br/>Atlanta, GA +<br/>Imaging and Audio Lab. +<br/>Texas Instruments +<br/>Dallas, TX +<br/>Center for Signal and Image Proc. +<br/>Georgia Inst. Of Technology +<br/>Atlanta, CA +</td></tr><tr><td>3504907a2e3c81d78e9dfe71c93ac145b1318f9c</td><td>Noname manuscript No. +<br/>(will be inserted by the editor) +<br/>Unconstrained Still/Video-Based Face Verification with Deep +<br/>Convolutional Neural Networks +<br/>Received: date / Accepted: date +</td></tr><tr><td>35b1c1f2851e9ac4381ef41b4d980f398f1aad68</td><td>Geometry Guided Convolutional Neural Networks for +<br/>Self-Supervised Video Representation Learning +</td></tr><tr><td>351c02d4775ae95e04ab1e5dd0c758d2d80c3ddd</td><td>ActionSnapping: Motion-based Video +<br/>Synchronization +<br/>Disney Research +</td></tr><tr><td>35e4b6c20756cd6388a3c0012b58acee14ffa604</td><td>Gender Classification in Large Databases +<br/>E. Ram´on-Balmaseda, J. Lorenzo-Navarro, and M. Castrill´on-Santana (cid:63) +<br/>Universidad de Las Palmas de Gran Canaria +<br/>SIANI +<br/>Spain +</td></tr><tr><td>357963a46dfc150670061dbc23da6ba7d6da786e</td><td></td></tr><tr><td>35f1bcff4552632419742bbb6e1927ef5e998eb4</td><td></td></tr><tr><td>35c973dba6e1225196566200cfafa150dd231fa8</td><td></td></tr><tr><td>35f084ddee49072fdb6e0e2e6344ce50c02457ef</td><td>A Bilinear Illumination Model +<br/>for Robust Face Recognition +<br/>The Harvard community has made this +<br/>article openly available. Please share how +<br/>this access benefits you. Your story matters +<br/>Citation +<br/>Machiraju. 2005. A bilinear illumination model for robust face +<br/>recognition. Proceedings of the Tenth IEEE International Conference +<br/>on Computer Vision: October 17-21, 2005, Beijing, China. 1177-1184. +<br/>Los Almamitos, C.A.: IEEE Computer Society. +<br/>Published Version +<br/>doi:10.1109/ICCV.2005.5 +<br/>Citable link +<br/>http://nrs.harvard.edu/urn-3:HUL.InstRepos:4238979 +<br/>Terms of Use +<br/><b></b><br/>repository, and is made available under the terms and conditions +<br/>applicable to Other Posted Material, as set forth at http:// +<br/>nrs.harvard.edu/urn-3:HUL.InstRepos:dash.current.terms-of- +<br/>use#LAA +</td></tr><tr><td>353a89c277cca3e3e4e8c6a199ae3442cdad59b5</td><td></td></tr><tr><td>352110778d2cc2e7110f0bf773398812fd905eb1</td><td>TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. X, NO. X, JUNE 2014 +<br/>Matrix Completion for Weakly-supervised +<br/>Multi-label Image Classification +</td></tr><tr><td>6964af90cf8ac336a2a55800d9c510eccc7ba8e1</td><td>Temporal Relational Reasoning in Videos +<br/>MIT CSAIL +</td></tr><tr><td>697b0b9630213ca08a1ae1d459fabc13325bdcbb</td><td></td></tr><tr><td>69d29012d17cdf0a2e59546ccbbe46fa49afcd68</td><td>Subspace clustering of dimensionality-reduced data +<br/>ETH Zurich, Switzerland +</td></tr><tr><td>69de532d93ad8099f4d4902c4cad28db958adfea</td><td></td></tr><tr><td>69526cdf6abbfc4bcd39616acde544568326d856</td><td>636 +<br/>[17] B. Moghaddam, T. Jebara, and A. Pentland, “Bayesian face recogni- +<br/>tion,” Pattern Recognit., vol. 33, no. 11, pp. 1771–1782, Nov. 2000. +<br/>[18] A. Nefian, “A hidden Markov model-based approach for face detection +<br/>and recognition,” Ph.D. dissertation, Dept. Elect. Comput. Eng. Elect. +<br/>Eng., Georgia Inst. Technol., Atlanta, 1999. +<br/>[19] P. J. Phillips et al., “Overview of the face recognition grand challenge,” +<br/>presented at the IEEE CVPR, San Diego, CA, Jun. 2005. +<br/>[20] H. T. Tanaka, M. Ikeda, and H. Chiaki, “Curvature-based face surface +<br/>recognition using spherical correlation-principal direction for curved +<br/>object recognition,” in Proc. Int. Conf. Automatic Face and Gesture +<br/>Recognition, 1998, pp. 372–377. +<br/>[21] M. Turk and A. Pentland, “Eigenfaces for recognition,” J. Cognit. Sci., +<br/>pp. 71–86, 1991. +<br/>[22] V. N. Vapnik, Statistical Learning Theory. New York: Wiley, 1998. +<br/>[23] W. Zhao, R. Chellappa, A. Rosenfeld, and P. Phillips, “Face recogni- +<br/>tion: A literature survey,” ACM Comput. Surveys, vol. 35, no. 44, pp. +<br/>399–458, 2003. +<br/>[24] W. Zhao, R. Chellappa, and P. J. Phillips, “Subspace linear discrimi- +<br/>nant analysis for face recognition,” UMD TR4009, 1999. +<br/>Face Verification Using Template Matching +</td></tr><tr><td>690d669115ad6fabd53e0562de95e35f1078dfbb</td><td>Progressive versus Random Projections for Compressive Capture of Images, +<br/>Lightfields and Higher Dimensional Visual Signals +<br/>MIT Media Lab +<br/>75 Amherst St, Cambridge, MA +<br/>MERL +<br/>201 Broadway, Cambridge MA +<br/>MIT Media Lab +<br/>75 Amherst St, Cambridge, MA +</td></tr><tr><td>69a9da55bd20ce4b83e1680fbc6be2c976067631</td><td></td></tr><tr><td>6974449ce544dc208b8cc88b606b03d95c8fd368</td><td></td></tr><tr><td>3cfbe1f100619a932ba7e2f068cd4c41505c9f58</td><td>A Realistic Simulation Tool for Testing Face Recognition +<br/>Systems under Real-World Conditions∗ +<br/>M. Correa, J. Ruiz-del-Solar, S. Parra-Tsunekawa, R. Verschae +<br/>Department of Electrical Engineering, Universidad de Chile +<br/>Advanced Mining Technology Center, Universidad de Chile +</td></tr><tr><td>3c03d95084ccbe7bf44b6d54151625c68f6e74d0</td><td></td></tr><tr><td>3cd7b15f5647e650db66fbe2ce1852e00c05b2e4</td><td></td></tr><tr><td>3ce2ecf3d6ace8d80303daf67345be6ec33b3a93</td><td></td></tr><tr><td>3c374cb8e730b64dacb9fbf6eb67f5987c7de3c8</td><td>Measuring Gaze Orientation for Human-Robot +<br/>Interaction +<br/>∗ CNRS; LAAS; 7 avenue du Colonel Roche, 31077 Toulouse Cedex, France +<br/>† Universit´e de Toulouse; UPS; LAAS-CNRS : F-31077 Toulouse, France +<br/>Introduction +<br/>In the context of Human-Robot interaction estimating gaze orientation brings +<br/>useful information about human focus of attention. This is a contextual infor- +<br/>mation : when you point something you usually look at it. Estimating gaze +<br/>orientation requires head pose estimation. There are several techniques to esti- +<br/>mate head pose from images, they are mainly based on training [3, 4] or on local +<br/>face features tracking [6]. The approach described here is based on local face +<br/>features tracking in image space using online learning, it is a mixed approach +<br/>since we track face features using some learning at feature level. It uses SURF +<br/>features [2] to guide detection and tracking. Such key features can be matched +<br/>between images, used for object detection or object tracking [10]. Several ap- +<br/>proaches work on fixed size images like training techniques which mainly work +<br/>on low resolution images because of computation costs whereas approaches based +<br/>on local features tracking work on high resolution images. Tracking face features +<br/>such as eyes, nose and mouth is a common problem in many applications such as +<br/>detection of facial expression or video conferencing [8] but most of those appli- +<br/>cations focus on front face images [9]. We developed an algorithm based on face +<br/>features tracking using a parametric model. First we need face detection, then +<br/>we detect face features in following order: eyes, mouth, nose. In order to achieve +<br/>full profile detection we use sets of SURF to learn what eyes, mouth and nose +<br/>look like once tracking is initialized. Once those sets of SURF are known they +<br/>are used to detect and track face features. SURF have a descriptor which is often +<br/>used to identify a key point and here we add some global geometry information +<br/>by using the relative position between key points. Then we use a particle filter to +<br/>track face features using those SURF based detectors, we compute the head pose +<br/>angles from features position and pass the results through a median filter. This +<br/>paper is organized as follows. Section 2 describes our modeling of visual features, +<br/>section 3 presents our tracking implementation. Section 4 presents results we get +<br/>with our implementation and future works in section 5. +<br/>2 Visual features +<br/>We use some basic properties of facial features to initialize our algorithm : eyes +<br/>are dark and circular, mouth is an horizontal dark line with a specific color,... +</td></tr><tr><td>3cb64217ca2127445270000141cfa2959c84d9e7</td><td></td></tr><tr><td>3cd5da596060819e2b156e8b3a28331ef633036b</td><td></td></tr><tr><td>3c56acaa819f4e2263638b67cea1ec37a226691d</td><td>Body Joint guided 3D Deep Convolutional +<br/>Descriptors for Action Recognition +</td></tr><tr><td>3c8da376576938160cbed956ece838682fa50e9f</td><td>Chapter 4 +<br/>Aiding Face Recognition with +<br/>Social Context Association Rule +<br/>based Re-Ranking +<br/>Humans are very efficient at recognizing familiar face images even in challenging condi- +<br/>tions. One reason for such capabilities is the ability to understand social context between +<br/>individuals. Sometimes the identity of the person in a photo can be inferred based on the +<br/>identity of other persons in the same photo, when some social context between them is +<br/>known. This chapter presents an algorithm to utilize the co-occurrence of individuals as +<br/>the social context to improve face recognition. Association rule mining is utilized to infer +<br/>multi-level social context among subjects from a large repository of social transactions. +<br/>The results are demonstrated on the G-album and on the SN-collection pertaining to 4675 +<br/>identities prepared by the authors from a social networking website. The results show that +<br/>association rules extracted from social context can be used to augment face recognition and +<br/>improve the identification performance. +<br/>4.1 +<br/>Introduction +<br/>Face recognition capabilities of humans have inspired several researchers to understand +<br/>the science behind it and use it in developing automated algorithms. Recently, it is also +<br/>argued that encoding social context among individuals can be leveraged for improved +<br/>automatic face recognition [175]. As shown in Figure 4.1, often times a person’s identity +<br/>can be inferred based on the identity of other persons in the same photo, when some social +<br/>context between them is known. A subject’s face in consumer photos generally co-occur +<br/>along with their socially relevant people. With the advent of social networking services, +<br/>the social context between individuals is readily available. Face recognition performance +<br/>105 +</td></tr><tr><td>56e885b9094391f7d55023a71a09822b38b26447</td><td>FREQUENCY DECODED LOCAL BINARY PATTERN +<br/>Face Retrieval using Frequency Decoded Local +<br/>Descriptor +</td></tr><tr><td>56a653fea5c2a7e45246613049fb16b1d204fc96</td><td>3287 +<br/>Quaternion Collaborative and Sparse Representation +<br/>With Application to Color Face Recognition +<br/>representation-based +</td></tr><tr><td>5666ed763698295e41564efda627767ee55cc943</td><td>Manuscript +<br/>Click here to download Manuscript: template.tex +<br/>Click here to view linked References +<br/>Noname manuscript No. +<br/>(will be inserted by the editor) +<br/>Relatively-Paired Space Analysis: Learning a Latent Common +<br/>Space from Relatively-Paired Observations +<br/>Received: date / Accepted: date +</td></tr><tr><td>5615d6045301ecbc5be35e46cab711f676aadf3a</td><td>Discriminatively Learned Hierarchical Rank Pooling Networks +<br/>Received: date / Accepted: date +</td></tr><tr><td>566038a3c2867894a08125efe41ef0a40824a090</td><td>978-1-4244-2354-5/09/$25.00 ©2009 IEEE +<br/>1945 +<br/>ICASSP 2009 +</td></tr><tr><td>56dca23481de9119aa21f9044efd7db09f618704</td><td>Riemannian Dictionary Learning and Sparse +<br/>Coding for Positive Definite Matrices +</td></tr><tr><td>516a27d5dd06622f872f5ef334313350745eadc3</td><td>> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < +<br/>1 +<br/>Fine-Grained Facial Expression Analysis Us- +<br/>ing Dimensional Emotion Model +<br/> +</td></tr><tr><td>51c3050fb509ca685de3d9ac2e965f0de1fb21cc</td><td>Fantope Regularization in Metric Learning +<br/>Marc T. Law +<br/>Sorbonne Universit´es, UPMC Univ Paris 06, UMR 7606, LIP6, F-75005, Paris, France +</td></tr><tr><td>51c7c5dfda47647aef2797ac3103cf0e108fdfb4</td><td>CS 395T: Celebrity Look-Alikes ∗ +</td></tr><tr><td>519f4eb5fe15a25a46f1a49e2632b12a3b18c94d</td><td>Non-Lambertian Reflectance Modeling and +<br/>Shape Recovery of Faces using Tensor Splines +</td></tr><tr><td>51528cdce7a92835657c0a616c0806594de7513b</td><td></td></tr><tr><td>5161e38e4ea716dcfb554ccb88901b3d97778f64</td><td>SSPP-DAN: DEEP DOMAIN ADAPTATION NETWORK FOR +<br/>FACE RECOGNITION WITH SINGLE SAMPLE PER PERSON +<br/>School of Computing, KAIST, Republic of Korea +</td></tr><tr><td>51d1a6e15936727e8dd487ac7b7fd39bd2baf5ee</td><td>JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 +<br/>A Fast and Accurate System for Face Detection, +<br/>Identification, and Verification +</td></tr><tr><td>5157dde17a69f12c51186ffc20a0a6c6847f1a29</td><td>Evolutionary Cost-sensitive Extreme Learning +<br/>Machine +<br/>1 +</td></tr><tr><td>51dc127f29d1bb076d97f515dca4cc42dda3d25b</td><td></td></tr><tr><td>3daafe6389d877fe15d8823cdf5ac15fd919676f</td><td>Human Action Localization +<br/>with Sparse Spatial Supervision +</td></tr><tr><td>3db75962857a602cae65f60f202d311eb4627b41</td><td></td></tr><tr><td>3d36f941d8ec613bb25e80fb8f4c160c1a2848df</td><td>Out-of-sample generalizations for supervised +<br/>manifold learning for classification +</td></tr><tr><td>3d5a1be4c1595b4805a35414dfb55716e3bf80d8</td><td>Hidden Two-Stream Convolutional Networks for +<br/>Action Recognition +</td></tr><tr><td>3d62b2f9cef997fc37099305dabff356d39ed477</td><td>Joint Face Alignment and 3D Face +<br/>Reconstruction with Application to Face +<br/>Recognition +</td></tr><tr><td>3dc522a6576c3475e4a166377cbbf4ba389c041f</td><td></td></tr><tr><td>3dd4d719b2185f7c7f92cc97f3b5a65990fcd5dd</td><td>Ensemble of Hankel Matrices for +<br/>Face Emotion Recognition +<br/>DICGIM, Universit´a degli Studi di Palermo, +<br/>V.le delle Scienze, Ed. 6, 90128 Palermo, Italy, +<br/>DRAFT +<br/>To appear in ICIAP 2015 +</td></tr><tr><td>3dda181be266950ba1280b61eb63ac11777029f9</td><td></td></tr><tr><td>3d6ee995bc2f3e0f217c053368df659a5d14d5b5</td><td></td></tr><tr><td>3dd906bc0947e56d2b7bf9530b11351bbdff2358</td><td></td></tr><tr><td>3d1af6c531ebcb4321607bcef8d9dc6aa9f0dc5a</td><td>1892 +<br/>Random Multispace Quantization as +<br/>an Analytic Mechanism for BioHashing +<br/>of Biometric and Random Identity Inputs +</td></tr><tr><td>3d6943f1573f992d6897489b73ec46df983d776c</td><td></td></tr><tr><td>3d94f81cf4c3a7307e1a976dc6cb7bf38068a381</td><td>3846 +<br/>Data-Dependent Label Distribution Learning +<br/>for Age Estimation +</td></tr><tr><td>5859774103306113707db02fe2dd3ac9f91f1b9e</td><td></td></tr><tr><td>5892f8367639e9c1e3cf27fdf6c09bb3247651ed</td><td>Estimating Missing Features to Improve Multimedia Information Retrieval +</td></tr><tr><td>5850aab97e1709b45ac26bb7d205e2accc798a87</td><td></td></tr><tr><td>587f81ae87b42c18c565694c694439c65557d6d5</td><td>DeepFace: Face Generation using Deep Learning +</td></tr><tr><td>580054294ca761500ada71f7d5a78acb0e622f19</td><td>1331 +<br/>A Subspace Model-Based Approach to Face +<br/>Relighting Under Unknown Lighting and Poses +</td></tr><tr><td>58081cb20d397ce80f638d38ed80b3384af76869</td><td>Embedded Real-Time Fall Detection Using Deep +<br/>Learning For Elderly Care +<br/>Samsung Research, Samsung Electronics +</td></tr><tr><td>58fa85ed57e661df93ca4cdb27d210afe5d2cdcd</td><td>Cancún Center, Cancún, México, December 4-8, 2016 +<br/>978-1-5090-4847-2/16/$31.00 ©2016 IEEE +<br/>4118 +</td></tr><tr><td>58bf72750a8f5100e0c01e55fd1b959b31e7dbce</td><td>PyramidBox: A Context-assisted Single Shot +<br/>Face Detector. +<br/>Baidu Inc. +</td></tr><tr><td>58542eeef9317ffab9b155579256d11efb4610f2</td><td>International Journal of Science and Research (IJSR) +<br/>ISSN (Online): 2319-7064 +<br/>Index Copernicus Value (2013): 6.14 | Impact Factor (2014): 5.611 +<br/>Face Recognition Revisited on Pose, Alignment, +<br/>Color, Illumination and Expression-PyTen +<br/>Computer Science, BIT Noida, India +</td></tr><tr><td>58823377757e7dc92f3b70a973be697651089756</td><td>Technical Report +<br/>UCAM-CL-TR-861 +<br/>ISSN 1476-2986 +<br/>Number 861 +<br/>Computer Laboratory +<br/>Automatic facial expression analysis +<br/>October 2014 +<br/>15 JJ Thomson Avenue +<br/>Cambridge CB3 0FD +<br/>United Kingdom +<br/>phone +44 1223 763500 +<br/>http://www.cl.cam.ac.uk/ +</td></tr><tr><td>58bb77dff5f6ee0fb5ab7f5079a5e788276184cc</td><td>Facial Expression Recognition with PCA and LBP +<br/>Features Extracting from Active Facial Patches +<br/> +</td></tr><tr><td>58cb1414095f5eb6a8c6843326a6653403a0ee17</td><td></td></tr><tr><td>677585ccf8619ec2330b7f2d2b589a37146ffad7</td><td>A flexible model for training action localization +<br/>with varying levels of supervision +</td></tr><tr><td>677477e6d2ba5b99633aee3d60e77026fb0b9306</td><td></td></tr><tr><td>6789bddbabf234f31df992a3356b36a47451efc7</td><td>Unsupervised Generation of Free-Form and +<br/>Parameterized Avatars +</td></tr><tr><td>675b2caee111cb6aa7404b4d6aa371314bf0e647</td><td>AVA: A Video Dataset of Spatio-temporally Localized Atomic Visual Actions +<br/>Carl Vondrick∗ +</td></tr><tr><td>679b72d23a9cfca8a7fe14f1d488363f2139265f</td><td></td></tr><tr><td>67484723e0c2cbeb936b2e863710385bdc7d5368</td><td>Anchor Cascade for Efficient Face Detection +</td></tr><tr><td>6742c0a26315d7354ab6b1fa62a5fffaea06da14</td><td>BAS AND SMITH: WHAT DOES 2D GEOMETRIC INFORMATION REALLY TELL US ABOUT 3D FACE SHAPE? +<br/>What does 2D geometric information +<br/>really tell us about 3D face shape? +</td></tr><tr><td>67a50752358d5d287c2b55e7a45cc39be47bf7d0</td><td></td></tr><tr><td>67ba3524e135c1375c74fe53ebb03684754aae56</td><td>978-1-5090-4117-6/17/$31.00 ©2017 IEEE +<br/>1767 +<br/>ICASSP 2017 +</td></tr><tr><td>6769cfbd85329e4815bb1332b118b01119975a95</td><td>Tied factor analysis for face recognition across +<br/>large pose changes +</td></tr><tr><td>0be43cf4299ce2067a0435798ef4ca2fbd255901</td><td>Title +<br/>A temporal latent topic model for facial expression recognition +<br/>Author(s) +<br/>Shang, L; Chan, KP +<br/>Citation +<br/>The 10th Asian Conference on Computer Vision (ACCV 2010), +<br/>Queenstown, New Zealand, 8-12 November 2010. In Lecture +<br/>Notes in Computer Science, 2010, v. 6495, p. 51-63 +<br/>Issued Date +<br/>2011 +<br/>URL +<br/>http://hdl.handle.net/10722/142604 +<br/>Rights +<br/>Creative Commons: Attribution 3.0 Hong Kong License +</td></tr><tr><td>0b2277a0609565c30a8ee3e7e193ce7f79ab48b0</td><td>944 +<br/>Cost-Sensitive Semi-Supervised Discriminant +<br/>Analysis for Face Recognition +</td></tr><tr><td>0ba64f4157d80720883a96a73e8d6a5f5b9f1d9b</td><td></td></tr><tr><td>0b605b40d4fef23baa5d21ead11f522d7af1df06</td><td>Label-Embedding for Attribute-Based Classification +<br/>a Computer Vision Group∗, XRCE, France +<br/>b LEAR†, INRIA, France +</td></tr><tr><td>0b0eb562d7341231c3f82a65cf51943194add0bb</td><td>> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < +<br/>Facial Image Analysis Based on Local Binary +<br/>Patterns: A Survey +<br/> +</td></tr><tr><td>0b3a146c474166bba71e645452b3a8276ac05998</td><td>Who’s in the Picture? +<br/>Berkeley, CA 94720 +<br/>Computer Science Division +<br/>U.C. Berkeley +</td></tr><tr><td>0b5bd3ce90bf732801642b9f55a781e7de7fdde0</td><td></td></tr><tr><td>0b0958493e43ca9c131315bcfb9a171d52ecbb8a</td><td>A Unified Neural Based Model for Structured Output Problems +<br/>Soufiane Belharbi∗1, Cl´ement Chatelain∗1, Romain H´erault∗1, and S´ebastien Adam∗2 +<br/>1LITIS EA 4108, INSA de Rouen, Saint ´Etienne du Rouvray 76800, France +<br/>2LITIS EA 4108, UFR des Sciences, Universit´e de Rouen, France. +<br/>April 13, 2015 +</td></tr><tr><td>0b20f75dbb0823766d8c7b04030670ef7147ccdd</td><td>1 +<br/>Feature selection using nearest attributes +</td></tr><tr><td>0b5a82f8c0ee3640503ba24ef73e672d93aeebbf</td><td>On Learning 3D Face Morphable Model +<br/>from In-the-wild Images +</td></tr><tr><td>0b174d4a67805b8796bfe86cd69a967d357ba9b6</td><td> Research Journal of Recent Sciences _________________________________________________ ISSN 2277-2502 +<br/> Vol. 3(4), 56-62, April (2014) +<br/>Res.J.Recent Sci. +</td></tr><tr><td>0ba449e312894bca0d16348f3aef41ca01872383</td><td></td></tr><tr><td>0b572a2b7052b15c8599dbb17d59ff4f02838ff7</td><td>Automatic Subspace Learning via Principal +<br/>Coefficients Embedding +</td></tr><tr><td>0ba99a709cd34654ac296418a4f41a9543928149</td><td></td></tr><tr><td>0b8c92463f8f5087696681fb62dad003c308ebe2</td><td>On Matching Sketches with Digital Face Images +<br/>in local +</td></tr><tr><td>0bc0f9178999e5c2f23a45325fa50300961e0226</td><td>Recognizing facial expressions from videos using Deep +<br/>Belief Networks +<br/>CS 229 Project +</td></tr><tr><td>0b3f354e6796ef7416bf6dde9e0779b2fcfabed2</td><td></td></tr><tr><td>93675f86d03256f9a010033d3c4c842a732bf661</td><td>Universit´edesSciencesetTechnologiesdeLilleEcoleDoctoraleSciencesPourl’ing´enieurUniversit´eLilleNord-de-FranceTHESEPr´esent´ee`al’Universit´edesSciencesetTechnologiesdeLillePourobtenirletitredeDOCTEURDEL’UNIVERSIT´ESp´ecialit´e:MicroetNanotechnologieParTaoXULocalizedgrowthandcharacterizationofsiliconnanowiresSoutenuele25Septembre2009Compositiondujury:Pr´esident:TuamiLASRIRapporteurs:ThierryBARONHenriMARIETTEExaminateurs:EricBAKKERSXavierWALLARTDirecteurdeth`ese:BrunoGRANDIDIER </td></tr><tr><td>936c7406de1dfdd22493785fc5d1e5614c6c2882</td><td>2012 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 762–772, +<br/>Montr´eal, Canada, June 3-8, 2012. c(cid:13)2012 Association for Computational Linguistics +<br/>762 +</td></tr><tr><td>93721023dd6423ab06ff7a491d01bdfe83db7754</td><td>ROBUST FACE ALIGNMENT USING CONVOLUTIONAL NEURAL +<br/>NETWORKS +<br/>Orange Labs, 4, Rue du Clos Courtel, 35512 Cesson-S´evign´e, France +<br/>Keywords: +<br/>Face alignment, Face registration, Convolutional Neural Networks. +</td></tr><tr><td>93cbb3b3e40321c4990c36f89a63534b506b6daf</td><td>IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART B: CYBERNETICS, VOL. 35, NO. 3, JUNE 2005 +<br/>477 +<br/>Learning From Examples in the Small Sample Case: +<br/>Face Expression Recognition +</td></tr><tr><td>944faf7f14f1bead911aeec30cc80c861442b610</td><td>Action Tubelet Detector for Spatio-Temporal Action Localization +</td></tr><tr><td>9458c518a6e2d40fb1d6ca1066d6a0c73e1d6b73</td><td>5967 +<br/>A Benchmark and Comparative Study of +<br/>Video-Based Face Recognition +<br/>on COX Face Database +</td></tr><tr><td>94aa8a3787385b13ee7c4fdd2b2b2a574ffcbd81</td><td></td></tr><tr><td>94325522c9be8224970f810554611d6a73877c13</td><td></td></tr><tr><td>9441253b638373a0027a5b4324b4ee5f0dffd670</td><td>A Novel Scheme for Generating Secure Face +<br/>Templates Using BDA +<br/>P.G. Student, Department of Computer Engineering, +<br/>Associate Professor, Department of Computer +<br/>MCERC, +<br/>Nashik (M.S.), India +</td></tr><tr><td>94ac3008bf6be6be6b0f5140a0bea738d4c75579</td><td></td></tr><tr><td>94a11b601af77f0ad46338afd0fa4ccbab909e82</td><td></td></tr><tr><td>0e8760fc198a7e7c9f4193478c0e0700950a86cd</td><td></td></tr><tr><td>0e50fe28229fea45527000b876eb4068abd6ed8c</td><td>Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence (IJCAI-17) +<br/>2936 +</td></tr><tr><td>0eff410cd6a93d0e37048e236f62e209bc4383d1</td><td>Anchorage Convention District +<br/>May 3-8, 2010, Anchorage, Alaska, USA +<br/>978-1-4244-5040-4/10/$26.00 ©2010 IEEE +<br/>4803 +</td></tr><tr><td>0ee737085af468f264f57f052ea9b9b1f58d7222</td><td>SiGAN: Siamese Generative Adversarial Network +<br/>for Identity-Preserving Face Hallucination +</td></tr><tr><td>0ee661a1b6bbfadb5a482ec643573de53a9adf5e</td><td>JOURNAL OF LATEX CLASS FILES, VOL. X, NO. X, MONTH YEAR +<br/>On the Use of Discriminative Cohort Score +<br/>Normalization for Unconstrained Face Recognition +</td></tr><tr><td>0e3840ea3227851aaf4633133dd3cbf9bbe89e5b</td><td></td></tr><tr><td>0e5dad0fe99aed6978c6c6c95dc49c6dca601e6a</td><td></td></tr><tr><td>0e2ea7af369dbcaeb5e334b02dd9ba5271b10265</td><td></td></tr><tr><td>0e7c70321462694757511a1776f53d629a1b38f3</td><td>NIST Special Publication 1136 +<br/>2012 Proceedings of the +<br/>Performance Metrics for Intelligent +<br/>Systems (PerMI ‘12) Workshop +<br/> +<br/>http://dx.doi.org/10.6028/NIST.SP.1136 +</td></tr><tr><td>6080f26675e44f692dd722b61905af71c5260af8</td><td></td></tr><tr><td>60d765f2c0a1a674b68bee845f6c02741a49b44e</td><td></td></tr><tr><td>60c24e44fce158c217d25c1bae9f880a8bd19fc3</td><td>Controllable Image-to-Video Translation: +<br/>A Case Study on Facial Expression Generation +<br/>MIT CSAIL +<br/>Wenbing Huang +<br/>Tencent AI Lab +<br/>MIT-Waston Lab +<br/>Tencent AI Lab +<br/>Tencent AI Lab +</td></tr><tr><td>60e2b9b2e0db3089237d0208f57b22a3aac932c1</td><td>Frankenstein: Learning Deep Face Representations +<br/>using Small Data +</td></tr><tr><td>60ce4a9602c27ad17a1366165033fe5e0cf68078</td><td>TECHNICAL NOTE +<br/>DIGITAL & MULTIMEDIA SCIENCES +<br/>J Forensic Sci, 2015 +<br/>doi: 10.1111/1556-4029.12800 +<br/>Available online at: onlinelibrary.wiley.com +<br/>Ph.D. +<br/>Combination of Face Regions in Forensic +<br/>Scenarios* +</td></tr><tr><td>6097ea6fd21a5f86a10a52e6e4dd5b78a436d5bf</td><td></td></tr><tr><td>60efdb2e204b2be6701a8e168983fa666feac1be</td><td>Int J Comput Vis +<br/>DOI 10.1007/s11263-017-1043-5 +<br/>Transferring Deep Object and Scene Representations for Event +<br/>Recognition in Still Images +<br/>Received: 31 March 2016 / Accepted: 1 September 2017 +<br/>© Springer Science+Business Media, LLC 2017 +</td></tr><tr><td>60824ee635777b4ee30fcc2485ef1e103b8e7af9</td><td>Cascaded Collaborative Regression for Robust Facial +<br/>Landmark Detection Trained using a Mixture of Synthetic and +<br/>Real Images with Dynamic Weighting +<br/>Life Member, IEEE, William Christmas, and Xiao-Jun Wu +</td></tr><tr><td>60643bdab1c6261576e6610ea64ea0c0b200a28d</td><td></td></tr><tr><td>60a20d5023f2bcc241eb9e187b4ddece695c2b9b</td><td>Invertible Nonlinear Dimensionality Reduction +<br/>via Joint Dictionary Learning +<br/>Department of Electrical and Computer Engineering +<br/>Technische Universit¨at M¨unchen, Germany +</td></tr><tr><td>60cdcf75e97e88638ec973f468598ae7f75c59b4</td><td>86 +<br/>Face Annotation Using Transductive +<br/>Kernel Fisher Discriminant +</td></tr><tr><td>60b3601d70f5cdcfef9934b24bcb3cc4dde663e7</td><td>SUBMITTED TO IEEE TRANS. ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE +<br/>Binary Gradient Correlation Patterns +<br/>for Robust Face Recognition +</td></tr><tr><td>34a41ec648d082270697b9ee264f0baf4ffb5c8d</td><td></td></tr><tr><td>341002fac5ae6c193b78018a164d3c7295a495e4</td><td>von Mises-Fisher Mixture Model-based Deep +<br/>learning: Application to Face Verification +</td></tr><tr><td>34ec83c8ff214128e7a4a4763059eebac59268a6</td><td>Action Anticipation By Predicting Future +<br/>Dynamic Images +<br/>Australian Centre for Robotic Vision, ANU, Canberra, Australia +</td></tr><tr><td>34b7e826db49a16773e8747bc8dfa48e344e425d</td><td></td></tr><tr><td>341ed69a6e5d7a89ff897c72c1456f50cfb23c96</td><td>DAGER: Deep Age, Gender and Emotion +<br/>Recognition Using Convolutional Neural +<br/>Networks +<br/>Computer Vision Lab, Sighthound Inc., Winter Park, FL +</td></tr><tr><td>340d1a9852747b03061e5358a8d12055136599b0</td><td>Audio-Visual Recognition System Insusceptible +<br/>to Illumination Variation over Internet Protocol +<br/> +</td></tr><tr><td>5a3da29970d0c3c75ef4cb372b336fc8b10381d7</td><td>CNN-based Real-time Dense Face Reconstruction +<br/>with Inverse-rendered Photo-realistic Face Images +</td></tr><tr><td>5a34a9bb264a2594c02b5f46b038aa1ec3389072</td><td>Label-Embedding for Image Classification +</td></tr><tr><td>5a4c6246758c522f68e75491eb65eafda375b701</td><td>978-1-4244-4296-6/10/$25.00 ©2010 IEEE +<br/>1118 +<br/>ICASSP 2010 +</td></tr><tr><td>5aad5e7390211267f3511ffa75c69febe3b84cc7</td><td>Driver Gaze Estimation +<br/>Without Using Eye Movement +<br/>MIT AgeLab +</td></tr><tr><td>5a029a0b0ae8ae7fc9043f0711b7c0d442bfd372</td><td></td></tr><tr><td>5a4ec5c79f3699ba037a5f06d8ad309fb4ee682c</td><td>Downloaded From: https://www.spiedigitallibrary.org/journals/Journal-of-Electronic-Imaging on 12/17/2017 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use +<br/>AutomaticageandgenderclassificationusingsupervisedappearancemodelAliMainaBukarHassanUgailDavidConnahAliMainaBukar,HassanUgail,DavidConnah,“Automaticageandgenderclassificationusingsupervisedappearancemodel,”J.Electron.Imaging25(6),061605(2016),doi:10.1117/1.JEI.25.6.061605. </td></tr><tr><td>5a7520380d9960ff3b4f5f0fe526a00f63791e99</td><td>The Indian Spontaneous Expression +<br/>Database for Emotion Recognition +</td></tr><tr><td>5fff61302adc65d554d5db3722b8a604e62a8377</td><td>Additive Margin Softmax for Face Verification +<br/>UESTC +<br/>Georgia Tech +<br/>UESTC +<br/>UESTC +</td></tr><tr><td>5fa6e4a23da0b39e4b35ac73a15d55cee8608736</td><td>IJCV special issue (Best papers of ECCV 2016) manuscript No. +<br/>(will be inserted by the editor) +<br/>RED-Net: +<br/>A Recurrent Encoder-Decoder Network for Video-based Face Alignment +<br/>Submitted: April 19 2017 / Revised: December 12 2017 +</td></tr><tr><td>5f871838710a6b408cf647aacb3b198983719c31</td><td>1716 +<br/>Locally Linear Regression for Pose-Invariant +<br/>Face Recognition +</td></tr><tr><td>5f64a2a9b6b3d410dd60dc2af4a58a428c5d85f9</td><td></td></tr><tr><td>5f344a4ef7edfd87c5c4bc531833774c3ed23542</td><td>c +</td></tr><tr><td>5fa0e6da81acece7026ac1bc6dcdbd8b204a5f0a</td><td></td></tr><tr><td>5f27ed82c52339124aa368507d66b71d96862cb7</td><td>Semi-supervised Learning of Classifiers: Theory, Algorithms +<br/>and Their Application to Human-Computer Interaction +<br/>This work has been partially funded by NSF Grant IIS 00-85980. +<br/>DRAFT +</td></tr><tr><td>5fa932be4d30cad13ea3f3e863572372b915bec8</td><td></td></tr><tr><td>5f5906168235613c81ad2129e2431a0e5ef2b6e4</td><td>Noname manuscript No. +<br/>(will be inserted by the editor) +<br/>A Unified Framework for Compositional Fitting of +<br/>Active Appearance Models +<br/>Received: date / Accepted: date +</td></tr><tr><td>5fc664202208aaf01c9b62da5dfdcd71fdadab29</td><td>arXiv:1504.05308v1 [cs.CV] 21 Apr 2015 +</td></tr><tr><td>5fa1724a79a9f7090c54925f6ac52f1697d6b570</td><td>Proceedings of the Workshop on Grammar and Lexicon: Interactions and Interfaces, +<br/>pages 41–47, Osaka, Japan, December 11 2016. +<br/>41 +</td></tr><tr><td>33aff42530c2fd134553d397bf572c048db12c28</td><td>From Emotions to Action Units with Hidden and Semi-Hidden-Task Learning +<br/>Universitat Pompeu Fabra +<br/>Centre de Visio per Computador +<br/>Universitat Pompeu Fabra +<br/>Barcelona +<br/>Barcelona +<br/>Barcelona +</td></tr><tr><td>33a1a049d15e22befc7ddefdd3ae719ced8394bf</td><td>FULL PAPER +<br/> International Journal of Recent Trends in Engineering, Vol 2, No. 1, November 2009 +<br/>An Efficient Approach to Facial Feature Detection +<br/>for Expression Recognition +<br/>S.P. Khandait1, P.D. Khandait2 and Dr.R.C.Thool2 +<br/>1Deptt. of Info.Tech., K.D.K.C.E., Nagpur, India +<br/> 2Deptt.of Electronics Engg., K.D.K.C.E., Nagpur, India, 2Deptt. of Info.Tech., SGGSIET, Nanded +</td></tr><tr><td>3399f8f0dff8fcf001b711174d29c9d4fde89379</td><td>Face R-CNN +<br/>Tencent AI Lab, China +</td></tr><tr><td>333aa36e80f1a7fa29cf069d81d4d2e12679bc67</td><td>Suggesting Sounds for Images +<br/>from Video Collections +<br/>1Computer Science Department, ETH Z¨urich, Switzerland +<br/>2Disney Research, Switzerland +</td></tr><tr><td>33792bb27ef392973e951ca5a5a3be4a22a0d0c6</td><td>Two-dimensional Whitening Reconstruction for +<br/>Enhancing Robustness of Principal Component +<br/>Analysis +</td></tr><tr><td>3328674d71a18ed649e828963a0edb54348ee598</td><td>IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART B: CYBERNETICS, VOL. 34, NO. 6, DECEMBER 2004 +<br/>2405 +<br/>A Face and Palmprint Recognition Approach Based +<br/>on Discriminant DCT Feature Extraction +</td></tr><tr><td>339937141ffb547af8e746718fbf2365cc1570c8</td><td>Facial Emotion Recognition in Real Time +</td></tr><tr><td>33aa980544a9d627f305540059828597354b076c</td><td></td></tr><tr><td>33ae696546eed070717192d393f75a1583cd8e2c</td><td></td></tr><tr><td>3352426a67eabe3516812cb66a77aeb8b4df4d1b</td><td>JOURNAL OF LATEX CLASS FILES, VOL. 4, NO. 5, APRIL 2015 +<br/>Joint Multi-view Face Alignment in the Wild +</td></tr><tr><td>334d6c71b6bce8dfbd376c4203004bd4464c2099</td><td>BICONVEX RELAXATION FOR SEMIDEFINITE PROGRAMMING IN +<br/>COMPUTER VISION +</td></tr><tr><td>33e20449aa40488c6d4b430a48edf5c4b43afdab</td><td>TRANSACTIONS ON AFFECTIVE COMPUTING +<br/>The Faces of Engagement: Automatic +<br/>Recognition of Student Engagement from Facial +<br/>Expressions +</td></tr><tr><td>333e7ad7f915d8ee3bb43a93ea167d6026aa3c22</td><td>This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication. +<br/>The final version of record is available at http://dx.doi.org/10.1109/TIFS.2014.2309851 +<br/>DRAFT +<br/>3D Assisted Face Recognition: Dealing With +<br/>Expression Variations +<br/> +</td></tr><tr><td>33403e9b4bbd913ae9adafc6751b52debbd45b0e</td><td></td></tr><tr><td>33ad23377eaead8955ed1c2b087a5e536fecf44e</td><td>Augmenting CRFs with Boltzmann Machine Shape Priors for Image Labeling +<br/>∗ indicates equal contribution +</td></tr><tr><td>05b8673d810fadf888c62b7e6c7185355ffa4121</td><td>(will be inserted by the editor) +<br/>A Comprehensive Survey to Face Hallucination +<br/>Received: date / Accepted: date +</td></tr><tr><td>05e658fed4a1ce877199a4ce1a8f8cf6f449a890</td><td></td></tr><tr><td>05ad478ca69b935c1bba755ac1a2a90be6679129</td><td>Attribute Dominance: What Pops Out? +<br/>Georgia Tech +</td></tr><tr><td>0562fc7eca23d47096472a1d42f5d4d086e21871</td><td></td></tr><tr><td>054738ce39920975b8dcc97e01b3b6cc0d0bdf32</td><td>Towards the Design of an End-to-End Automated +<br/>System for Image and Video-based Recognition +</td></tr><tr><td>05e03c48f32bd89c8a15ba82891f40f1cfdc7562</td><td>Scalable Robust Principal Component +<br/>Analysis using Grassmann Averages +</td></tr><tr><td>050fdbd2e1aa8b1a09ed42b2e5cc24d4fe8c7371</td><td>Contents +<br/>Scale Space and PDE Methods +<br/>Spatio-Temporal Scale Selection in Video Data . . . . . . . . . . . . . . . . . . . . . +<br/>Dynamic Texture Recognition Using Time-Causal Spatio-Temporal +<br/>Scale-Space Filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . +<br/>Corner Detection Using the Affine Morphological Scale Space . . . . . . . . . . . +<br/>Luis Alvarez +<br/>Nonlinear Spectral Image Fusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . +<br/>Martin Benning, Michael Möller, Raz Z. Nossek, Martin Burger, +<br/>Daniel Cremers, Guy Gilboa, and Carola-Bibiane Schönlieb +<br/>16 +<br/>29 +<br/>41 +<br/>Tubular Structure Segmentation Based on Heat Diffusion. . . . . . . . . . . . . . . +<br/>54 +<br/>Fang Yang and Laurent D. Cohen +<br/>Analytic Existence and Uniqueness Results for PDE-Based Image +<br/>Reconstruction with the Laplacian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . +<br/>Laurent Hoeltgen, Isaac Harris, Michael Breuß, and Andreas Kleefeld +<br/>Combining Contrast Invariant L1 Data Fidelities with Nonlinear +<br/>Spectral Image Decomposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . +<br/>Leonie Zeune, Stephan A. van Gils, Leon W.M.M. Terstappen, +<br/>and Christoph Brune +<br/>An Efficient and Stable Two-Pixel Scheme for 2D +<br/>Forward-and-Backward Diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . +<br/>Martin Welk and Joachim Weickert +<br/>66 +<br/>80 +<br/>94 +<br/>Restoration and Reconstruction +<br/>Blind Space-Variant Single-Image Restoration of Defocus Blur. . . . . . . . . . . +<br/>109 +<br/>Leah Bar, Nir Sochen, and Nahum Kiryati +<br/>Denoising by Inpainting. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . +<br/>121 +<br/>Robin Dirk Adam, Pascal Peter, and Joachim Weickert +<br/>Stochastic Image Reconstruction from Local Histograms +<br/>of Gradient Orientation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . +<br/>Agnès Desolneux and Arthur Leclaire +<br/>133 +</td></tr><tr><td>056294ff40584cdce81702b948f88cebd731a93e</td><td></td></tr><tr><td>052880031be0a760a5b606b2ad3d22f237e8af70</td><td>Datasets on object manipulation and interaction: a survey +</td></tr><tr><td>05ea7930ae26165e7e51ff11b91c7aa8d7722002</td><td>Learning And-Or Model to Represent Context and +<br/>Occlusion for Car Detection and Viewpoint Estimation +</td></tr><tr><td>051a84f0e39126c1ebeeb379a405816d5d06604d</td><td>Cogn Comput (2009) 1:257–267 +<br/>DOI 10.1007/s12559-009-9018-7 +<br/>Biometric Recognition Performing in a Bioinspired System +<br/>Joan Fa`bregas Æ Marcos Faundez-Zanuy +<br/>Published online: 20 May 2009 +<br/>Ó Springer Science+Business Media, LLC 2009 +</td></tr><tr><td>05f4d907ee2102d4c63a3dc337db7244c570d067</td><td></td></tr><tr><td>05a7be10fa9af8fb33ae2b5b72d108415519a698</td><td>Multilayer and Multimodal Fusion of Deep Neural Networks +<br/>for Video Classification +<br/>NVIDIA +</td></tr><tr><td>050a149051a5d268fcc5539e8b654c2240070c82</td><td>MAGISTERSKÉ A DOKTORSKÉSTUDIJNÍ PROGRAMY31. 5. 2018SBORNÍKSTUDENTSKÁ VĚDECKÁ KONFERENCE </td></tr><tr><td>0580edbd7865414c62a36da9504d1169dea78d6f</td><td>Baseline CNN structure analysis for facial expression recognition +</td></tr><tr><td>05e96d76ed4a044d8e54ef44dac004f796572f1a</td><td></td></tr><tr><td>9d839dfc9b6a274e7c193039dfa7166d3c07040b</td><td>Augmented Faces +<br/>1ETH Z¨urich +<br/>2Kooaba AG +<br/>3K.U. Leuven +</td></tr><tr><td>9d60ad72bde7b62be3be0c30c09b7d03f9710c5f</td><td>A Survey: Face Recognition Techniques +<br/>Assistant Professor, ITM GOI +<br/>M Tech, ITM GOI +<br/>face +<br/>video +<br/>(Eigen +<br/>passport-verification, +</td></tr><tr><td>9cfb3a68fb10a59ec2a6de1b24799bf9154a8fd1</td><td></td></tr><tr><td>9ca7899338129f4ba6744f801e722d53a44e4622</td><td>Deep Neural Networks Regularization for Structured +<br/>Output Prediction +<br/>Soufiane Belharbi∗ +<br/>INSA Rouen, LITIS +<br/>76000 Rouen, France +<br/>INSA Rouen, LITIS +<br/>76000 Rouen, France +<br/>INSA Rouen, LITIS +<br/>76000 Rouen, France +<br/>INSA Rouen, LITIS +<br/>76000 Rouen, France +<br/>Normandie Univ, UNIROUEN, UNIHAVRE, +<br/>Normandie Univ, UNIROUEN, UNIHAVRE, +<br/>Normandie Univ, UNIROUEN, UNIHAVRE, +<br/>Normandie Univ, UNIROUEN, UNIHAVRE, +</td></tr><tr><td>9c1664f69d0d832e05759e8f2f001774fad354d6</td><td>Action representations in robotics: A +<br/>taxonomy and systematic classification +<br/>Journal Title +<br/>XX(X):1–32 +<br/>c(cid:13)The Author(s) 2016 +<br/>Reprints and permission: +<br/>sagepub.co.uk/journalsPermissions.nav +<br/>DOI: 10.1177/ToBeAssigned +<br/>www.sagepub.com/ +</td></tr><tr><td>9c065dfb26ce280610a492c887b7f6beccf27319</td><td>Learning from Video and Text via Large-Scale Discriminative Clustering +<br/>1 ´Ecole Normale Sup´erieure +<br/>2Inria +<br/>3CIIRC +</td></tr><tr><td>02601d184d79742c7cd0c0ed80e846d95def052e</td><td>Graphical Representation for Heterogeneous +<br/>Face Recognition +</td></tr><tr><td>02cc96ad997102b7c55e177ac876db3b91b4e72c</td><td>MuseumVisitors: a dataset for pedestrian and group detection, gaze estimation +<br/>and behavior understanding +</td></tr><tr><td>02fda07735bdf84554c193811ba4267c24fe2e4a</td><td>Illumination Invariant Face Recognition +<br/>Using Near-Infrared Images +</td></tr><tr><td>02dd0af998c3473d85bdd1f77254ebd71e6158c6</td><td>PPP: Joint Pointwise and Pairwise Image Label Prediction +<br/>1Department of Computer Science, Arizona State Univerity +<br/>2Yahoo Research +</td></tr><tr><td>029317f260b3303c20dd58e8404a665c7c5e7339</td><td>1276 +<br/>Character Identification in Feature-Length Films +<br/>Using Global Face-Name Matching +<br/>and Yeh-Min Huang, Member, IEEE +</td></tr><tr><td>0273414ba7d56ab9ff894959b9d46e4b2fef7fd0</td><td>Photographic home styles in Congress: a +<br/>computer vision approach∗ +<br/>December 1, 2016 +</td></tr><tr><td>02e133aacde6d0977bca01ffe971c79097097b7f</td><td></td></tr><tr><td>02567fd428a675ca91a0c6786f47f3e35881bcbd</td><td>ACCEPTED BY IEEE TIP +<br/>Deep Label Distribution Learning +<br/>With Label Ambiguity +</td></tr><tr><td>029b53f32079063047097fa59cfc788b2b550c4b</td><td></td></tr><tr><td>02bd665196bd50c4ecf05d6852a4b9ba027cd9d0</td><td></td></tr><tr><td>026b5b8062e5a8d86c541cfa976f8eee97b30ab8</td><td>MDLFace: Memorability Augmented Deep Learning for Video Face Recognition +<br/>IIIT-Delhi, India +</td></tr><tr><td>0278acdc8632f463232e961563e177aa8c6d6833</td><td>Selective Transfer Machine for Personalized +<br/>Facial Expression Analysis +<br/>1 INTRODUCTION +<br/>Index Terms—Facial expression analysis, personalization, domain adaptation, transfer learning, support vector machine (SVM) +<br/>A UTOMATIC facial AU detection confronts a number of +</td></tr><tr><td>02c993d361dddba9737d79e7251feca026288c9c</td><td></td></tr><tr><td>a46283e90bcdc0ee35c680411942c90df130f448</td><td></td></tr><tr><td>a4a5ad6f1cc489427ac1021da7d7b70fa9a770f2</td><td>Yudistira and Kurita EURASIP Journal on Image and Video +<br/>Processing (2017) 2017:85 +<br/>DOI 10.1186/s13640-017-0235-9 +<br/>EURASIP Journal on Image +<br/>and Video Processing +<br/>RESEARCH +<br/>Open Access +<br/>Gated spatio and temporal convolutional +<br/>neural network for activity recognition: +<br/>towards gated multimodal deep learning +</td></tr><tr><td>a40f8881a36bc01f3ae356b3e57eac84e989eef0</td><td>End-to-end semantic face segmentation with conditional +<br/>random fields as convolutional, recurrent and adversarial +<br/>networks +</td></tr><tr><td>a44590528b18059b00d24ece4670668e86378a79</td><td>Learning the Hierarchical Parts of Objects by Deep +<br/>Non-Smooth Nonnegative Matrix Factorization +</td></tr><tr><td>a4c430b7d849a8f23713dc283794d8c1782198b2</td><td>Video Concept Embedding +<br/>1. Introduction +<br/>In the area of natural language processing, there has been +<br/>much success in learning distributed representations for +<br/>words as vectors. Doing so has an advantage over using +<br/>simple labels, or a one-hot coding scheme for representing +<br/>individual words. In learning distributed vector representa- +<br/>tions for words, we manage to capture semantic relatedness +<br/>of words in vector distance. For example, the word vector +<br/>for ”car” and ”road” should end up being closer together in +<br/>the vector space representation than ”car” and ”penguin”. +<br/>This has been very useful in NLP areas of machine transla- +<br/>tion and semantic understanding. +<br/>In the computer vision domain, video understanding is a +<br/>very important topic. +<br/>It is made hard due to the large +<br/>amount of high dimensional data in videos. One strategy +<br/>to address this is to summarize a video into concepts (eg. +<br/>running, climbing, cooking). This allows us to represent a +<br/>video in a very natural way to humans, such as a sequence +<br/>of semantic events. However this has the same shortcom- +<br/>ings that one-hot coding of words have. +<br/>The goal of this project is to find a meaningful way to em- +<br/>bed video concepts into a vector space. The hope would +<br/>be to capture semantic relatedness of concepts in a vector +<br/>representation, essentially doing for videos what word2vec +<br/>did for text. Having a vector representation for video con- +<br/>cepts would help in areas such as semantic video retrieval +<br/>and video classification, as it would provide a statistically +<br/>meaningful and robust way of representing videos as lower +<br/>dimensional vectors. An interesting thing would be to ob- +<br/>serve if such a vector representation would result in ana- +<br/>logical reasoning using simple vector arithmetic. +<br/>Figure 1 shows an example of concepts detected at differ- +<br/>ent snapshots in the same video. For example, consider +<br/>the scenario where the concepts Kicking a ball, Soccer and +<br/>Running are detected in the three snapshots respectively +<br/>(from left to right). Since, these snapshots belong in the +<br/>same video, we expect that these concepts are semantically +<br/>similar and that they should lie close in the resulting em- +<br/>bedding space. The aim of this project is to find a vector +<br/>space embedding for the space of concepts such that vector +<br/>representations for semantically similar concepts (in this +<br/>Figure 1. Example snapshots from the same video +<br/>case, Running, Kicking and Soccer) lie in the vicinity of +<br/>each other. +<br/>2. Related Work +<br/>(Mikolov et al., 2013a) introduces the popular skip-gram +<br/>model to learn distributed representations of words from +<br/>very large linguistic datasets. Specifically, it uses each +<br/>word as an input to a log-linear classifier and predict words +<br/>within a certain range before and after the current word in +<br/>the dataset. +<br/>(Mikolov et al., 2013b) extends this model +<br/>to learn representations for phrases, in addition to words, +<br/>and also improve the quality of vectors and training speed. +<br/>These works also show that the skip-gram model exhibits +<br/>a linear structure that enables it to perform reasoning using +<br/>basic vector arithmetic. The skip-gram model from these +<br/>works is the basis of our model in learning representations +<br/>for concepts. +<br/>(Le & Mikolov, 2014) extends the concept of word vectors +<br/>to sentences and paragraphs. Their approach is more in- +<br/>volved than a simple bag of words approach, in that it tries +<br/>to capture the nature of the words in the paragraph. They +<br/>construct the paragraph vector in such a way that it can be +<br/>used to predict the word vectors that are contained inside +<br/>the paragraph. They do this by first learning word vectors, +<br/>such that the probability of a word vector given its context +<br/>is maximized. To learn paragraph vectors, the paragraph +<br/>is essentially treated as a word, and the words it contains +<br/>become the context. This provides a key insight in how +<br/>a set of concept vectors can be used together to provide a +<br/>more meaningful vector representation for videos, which +<br/>can then be used for retrieval. +<br/>(Hu et al.) utilizes structured knowledge in the data to learn +<br/>distributed representations that improve semantic related- +</td></tr><tr><td>a4cc626da29ac48f9b4ed6ceb63081f6a4b304a2</td><td></td></tr><tr><td>a4f37cfdde3af723336205b361aefc9eca688f5c</td><td>Recent Advances +<br/>in Face Recognition +</td></tr><tr><td>a30869c5d4052ed1da8675128651e17f97b87918</td><td>Fine-Grained Comparisons with Attributes +</td></tr><tr><td>a3ebacd8bcbc7ddbd5753935496e22a0f74dcf7b</td><td>First International Workshop on Adaptive Shot Learning +<br/>for Gesture Understanding and Production +<br/>ASL4GUP 2017 +<br/>Held in conjunction with IEEE FG 2017, in May 30, 2017, +<br/>Washington DC, USA +</td></tr><tr><td>a3d8b5622c4b9af1f753aade57e4774730787a00</td><td>Pose-Aware Person Recognition +<br/>Anoop Namboodiri (cid:63) +<br/>(cid:63) CVIT, IIIT Hyderabad, India +<br/>† Facebook AI Research +</td></tr><tr><td>a3017bb14a507abcf8446b56243cfddd6cdb542b</td><td>Face Localization and Recognition in Varied +<br/>Expressions and Illumination +<br/>Hui-Yu Huang, Shih-Hang Hsu +<br/> +</td></tr><tr><td>a378fc39128107815a9a68b0b07cffaa1ed32d1f</td><td>Determining a Suitable Metric When using Non-negative Matrix Factorization∗ +<br/>Computer Vision Center, Dept. Inform`atica +<br/>Universitat Aut`onoma de Barcelona +<br/>08193 Bellaterra, Barcelona, Spain +</td></tr><tr><td>a34d75da87525d1192bda240b7675349ee85c123</td><td>Naive-Deep Face Recognition: Touching the Limit of LFW Benchmark or Not? +<br/>Face++, Megvii Inc. +<br/>Face++, Megvii Inc. +<br/>Face++, Megvii Inc. +</td></tr><tr><td>a3f69a073dcfb6da8038607a9f14eb28b5dab2db</td><td>Proceedings of the Twenty-Seventh International Joint Conference on Artificial Intelligence (IJCAI-18) +<br/>1184 +</td></tr><tr><td>a3f78cc944ac189632f25925ba807a0e0678c4d5</td><td>Action Recognition in Realistic Sports Videos +</td></tr><tr><td>a33f20773b46283ea72412f9b4473a8f8ad751ae</td><td></td></tr><tr><td>a3a6a6a2eb1d32b4dead9e702824375ee76e3ce7</td><td>Multiple Local Curvature Gabor Binary +<br/>Patterns for Facial Action Recognition +<br/>Signal Processing Laboratory (LTS5), +<br/>´Ecole Polytechnique F´ed´erale de Lausanne, Switzerland +</td></tr><tr><td>a32c5138c6a0b3d3aff69bcab1015d8b043c91fb</td><td>Downloaded From: https://www.spiedigitallibrary.org/journals/Journal-of-Electronic-Imaging on 9/19/2018 +<br/>Terms of Use: https://www.spiedigitallibrary.org/terms-of-use +<br/>Videoredaction:asurveyandcomparisonofenablingtechnologiesShaganSahAmeyaShringiRaymondPtuchaAaronBurryRobertLoceShaganSah,AmeyaShringi,RaymondPtucha,AaronBurry,RobertLoce,“Videoredaction:asurveyandcomparisonofenablingtechnologies,”J.Electron.Imaging26(5),051406(2017),doi:10.1117/1.JEI.26.5.051406. </td></tr><tr><td>a3d78bc94d99fdec9f44a7aa40c175d5a106f0b9</td><td>Recognizing Violence in Movies +<br/>CIS400/401 Project Final Report +<br/>Univ. of Pennsylvania +<br/>Philadelphia, PA +<br/>Univ. of Pennsylvania +<br/>Philadelphia, PA +<br/>Ben Sapp +<br/>Univ. of Pennsylvania +<br/>Philadelphia, PA +<br/>Univ. of Pennsylvania +<br/>Philadelphia, PA +</td></tr><tr><td>a3eab933e1b3db1a7377a119573ff38e780ea6a3</td><td>978-1-4244-4296-6/10/$25.00 ©2010 IEEE +<br/>838 +<br/>ICASSP 2010 +</td></tr><tr><td>a35d3ba191137224576f312353e1e0267e6699a1</td><td>Increasing security in DRM systems +<br/>through biometric authentication. +<br/>ecuring the exchange +<br/>of intellectual property +<br/>and providing protection +<br/>to multimedia contents in +<br/>distribution systems have enabled the +<br/>advent of digital rights management +<br/>(DRM) systems [5], [14], [21], [47], +<br/>[51], [53]. Rights holders should be able to +<br/>license, monitor, and track the usage of rights +<br/>in a dynamic digital trading environment, espe- +<br/>cially in the near future when universal multimedia +<br/>access (UMA) becomes a reality, and any multimedia +<br/>content will be available anytime, anywhere. In such +<br/>DRM systems, encryption algorithms, access control, +<br/>key management strategies, identification and tracing +<br/>of contents, or copy control will play a prominent role +<br/>to supervise and restrict access to multimedia data, +<br/>avoiding unauthorized or fraudulent operations. +<br/>A key component of any DRM system, also known +<br/>as intellectual property management and protection +<br/>(IPMP) systems in the MPEG-21 framework, is user +<br/>authentication to ensure that +<br/>only those with specific rights are +<br/>able to access the digital informa- +<br/>tion. It is here that biometrics can +<br/>play an essential role, reinforcing securi- +<br/>ty at all stages where customer authentica- +<br/>tion is needed. The ubiquity of users and +<br/>devices, where the same user might want to +<br/>access to multimedia contents from different +<br/>environments (home, car, work, jogging, etc.) and +<br/>also from different devices or media (CD, DVD, +<br/>home computer, laptop, PDA, 2G/3G mobile phones, +<br/>game consoles, etc.) strengthens the need for reliable +<br/>and universal authentication of users. +<br/>Classical user authentication systems have been +<br/>based in something that you have (like a key, an identi- +<br/>fication card, etc.) and/or something that you know +<br/>(like a password, or a PIN). With biometrics, a new +<br/>user authentication paradigm is added: something that +<br/>you are (e.g., fingerprints or face) or something that +<br/>you do or produce (e.g., handwritten signature or +<br/>50 +<br/>IEEE SIGNAL PROCESSING MAGAZINE +<br/>1053-5888/04/$20.00©2004IEEE +<br/>MARCH 2004 +</td></tr><tr><td>b558be7e182809f5404ea0fcf8a1d1d9498dc01a</td><td>Bottom-up and top-down reasoning with convolutional latent-variable models +<br/>UC Irvine +<br/>UC Irvine +</td></tr><tr><td>b5fc4f9ad751c3784eaf740880a1db14843a85ba</td><td>SIViP (2007) 1:225–237 +<br/>DOI 10.1007/s11760-007-0016-5 +<br/>ORIGINAL PAPER +<br/>Significance of image representation for face verification +<br/>Received: 29 August 2006 / Revised: 28 March 2007 / Accepted: 28 March 2007 / Published online: 1 May 2007 +<br/>© Springer-Verlag London Limited 2007 +</td></tr><tr><td>b562def2624f59f7d3824e43ecffc990ad780898</td><td></td></tr><tr><td>b5160e95192340c848370f5092602cad8a4050cd</td><td>IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY, TO APPEAR +<br/>Video Classification With CNNs: Using The Codec +<br/>As A Spatio-Temporal Activity Sensor +</td></tr><tr><td>b52c0faba5e1dc578a3c32a7f5cfb6fb87be06ad</td><td>Journal of Applied Research and +<br/>Technology +<br/>ISSN: 1665-6423 +<br/>Centro de Ciencias Aplicadas y +<br/>Desarrollo Tecnológico +<br/>México +<br/> +<br/>Hussain Shah, Jamal; Sharif, Muhammad; Raza, Mudassar; Murtaza, Marryam; Ur-Rehman, Saeed +<br/>Robust Face Recognition Technique under Varying Illumination +<br/>Journal of Applied Research and Technology, vol. 13, núm. 1, febrero, 2015, pp. 97-105 +<br/>Centro de Ciencias Aplicadas y Desarrollo Tecnológico +<br/>Distrito Federal, México +<br/>Available in: http://www.redalyc.org/articulo.oa?id=47436895009 +<br/> How to cite +<br/> Complete issue +<br/> More information about this article +<br/> Journal's homepage in redalyc.org +<br/>Scientific Information System +<br/>Network of Scientific Journals from Latin America, the Caribbean, Spain and Portugal +<br/>Non-profit academic project, developed under the open access initiative +</td></tr><tr><td>b52886610eda6265a2c1aaf04ce209c047432b6d</td><td>Microexpression Identification and Categorization +<br/>using a Facial Dynamics Map +</td></tr><tr><td>b5857b5bd6cb72508a166304f909ddc94afe53e3</td><td>SSIG and IRISA at Multimodal Person Discovery +<br/>1Department of Computer Science, Universidade Federal de Minas Gerais, Belo Horizonte, Brazil +<br/>2IRISA & Inria Rennes , CNRS, Rennes, France +</td></tr><tr><td>b59f441234d2d8f1765a20715e227376c7251cd7</td><td></td></tr><tr><td>b51e3d59d1bcbc023f39cec233f38510819a2cf9</td><td>CBMM Memo No. 003 +<br/>March 27, 2014 +<br/>Can a biologically-plausible hierarchy effectively +<br/>replace face detection, alignment, and +<br/>recognition pipelines? +<br/>by +</td></tr><tr><td>b54c477885d53a27039c81f028e710ca54c83f11</td><td>1201 +<br/>Semi-Supervised Kernel Mean Shift Clustering +</td></tr><tr><td>b2a0e5873c1a8f9a53a199eecae4bdf505816ecb</td><td>Hybrid VAE: Improving Deep Generative Models +<br/>using Partial Observations +<br/>Snap Research +<br/>Microsoft Research +</td></tr><tr><td>b2b535118c5c4dfcc96f547274cdc05dde629976</td><td>JOURNAL OF IEEE TRANSACTIONS ON AFFECTIVE COMPUTING, VOL. XX, NO. X, XXX 2017 +<br/>Automatic Recognition of Facial Displays of +<br/>Unfelt Emotions +<br/>Escalera, Xavier Bar´o, Sylwia Hyniewska, Member, IEEE, J¨uri Allik, +</td></tr><tr><td>b235b4ccd01a204b95f7408bed7a10e080623d2e</td><td>Regularizing Flat Latent Variables with Hierarchical Structures +</td></tr><tr><td>b2c25af8a8e191c000f6a55d5f85cf60794c2709</td><td>Noname manuscript No. +<br/>(will be inserted by the editor) +<br/>A Novel Dimensionality Reduction Technique based on +<br/>Kernel Optimization Through Graph Embedding +<br/>N. Vretos, A. Tefas and I. Pitas +<br/>the date of receipt and acceptance should be inserted later +</td></tr><tr><td>d904f945c1506e7b51b19c99c632ef13f340ef4c</td><td>A scalable 3D HOG model for fast object detection and viewpoint estimation +<br/>KU Leuven, ESAT/PSI - iMinds +<br/>Kasteelpark Arenberg 10 B-3001 Leuven, Belgium +</td></tr><tr><td>d94d7ff6f46ad5cab5c20e6ac14c1de333711a0c</td><td>978-1-5090-4117-6/17/$31.00 ©2017 IEEE +<br/>3031 +<br/>ICASSP 2017 +</td></tr><tr><td>d9739d1b4478b0bf379fe755b3ce5abd8c668f89</td><td></td></tr><tr><td>d9318c7259e394b3060b424eb6feca0f71219179</td><td>406 +<br/>Face Matching and Retrieval Using Soft Biometrics +</td></tr><tr><td>d9a1dd762383213741de4c1c1fd9fccf44e6480d</td><td></td></tr><tr><td>d9c4b1ca997583047a8721b7dfd9f0ea2efdc42c</td><td>Learning Inference Models for Computer Vision +</td></tr><tr><td>aca232de87c4c61537c730ee59a8f7ebf5ecb14f</td><td>EBGM VS SUBSPACE PROJECTION FOR FACE RECOGNITION +<br/>19.5 Km Markopoulou Avenue, P.O. Box 68, Peania, Athens, Greece +<br/>Athens Information Technology +<br/>Keywords: +<br/>Human-Machine Interfaces, Computer Vision, Face Recognition. +</td></tr><tr><td>ac6a9f80d850b544a2cbfdde7002ad5e25c05ac6</td><td>779 +<br/>Privacy-Protected Facial Biometric Verification +<br/>Using Fuzzy Forest Learning +</td></tr><tr><td>aca273a9350b10b6e2ef84f0e3a327255207d0f5</td><td></td></tr><tr><td>ac0d3f6ed5c42b7fc6d7c9e1a9bb80392742ad5e</td><td></td></tr><tr><td>ac820d67b313c38b9add05abef8891426edd5afb</td><td></td></tr><tr><td>ac26166857e55fd5c64ae7194a169ff4e473eb8b</td><td>Personalized Age Progression with Bi-level +<br/>Aging Dictionary Learning +</td></tr><tr><td>ac8441e30833a8e2a96a57c5e6fede5df81794af</td><td>IEEE TRANSACTIONS ON IMAGE PROCESSING +<br/>Hierarchical Representation Learning for Kinship +<br/>Verification +</td></tr><tr><td>acb83d68345fe9a6eb9840c6e1ff0e41fa373229</td><td>Kernel Methods in Computer Vision: +<br/>Object Localization, Clustering, +<br/>and Taxonomy Discovery +<br/>vorgelegt von +<br/>Matthew Brian Blaschko, M.S. +<br/>aus La Jolla +<br/>Von der Fakult¨at IV - Elektrotechnik und Informatik +<br/>der Technischen Universit¨at Berlin +<br/>zur Erlangung des akademischen Grades +<br/>Doktor der Naturwissenschaften +<br/>Dr. rer. nat. +<br/>genehmigte Dissertation +<br/>Promotionsausschuß: +<br/>Vorsitzender: Prof. Dr. O. Hellwich +<br/>Berichter: Prof. Dr. T. Hofmann +<br/>Berichter: Prof. Dr. K.-R. M¨uller +<br/>Berichter: Prof. Dr. B. Sch¨olkopf +<br/>Tag der wissenschaftlichen Aussprache: 23.03.2009 +<br/>Berlin 2009 +<br/>D83 +</td></tr><tr><td>adf7ccb81b8515a2d05fd3b4c7ce5adf5377d9be</td><td>Apprentissage de métrique appliqué à la +<br/>détection de changement de page Web et +<br/>aux attributs relatifs +<br/>thieu Cord* +<br/>* Sorbonne Universités, UPMC Univ Paris 06, UMR 7606, LIP6, F-75005, Paris, +<br/>France +<br/>RÉSUMÉ. Nous proposons dans cet article un nouveau schéma d’apprentissage de métrique. +<br/>Basé sur l’exploitation de contraintes qui impliquent des quadruplets d’images, notre approche +<br/>vise à modéliser des relations sémantiques de similarités riches ou complexes. Nous étudions +<br/>comment ce schéma peut être utilisé dans des contextes tels que la détection de régions impor- +<br/>tantes dans des pages Web ou la reconnaissance à partir d’attributs relatifs. +</td></tr><tr><td>ada73060c0813d957576be471756fa7190d1e72d</td><td>VRPBench: A Vehicle Routing Benchmark Tool +<br/>October 19, 2016 +</td></tr><tr><td>adaf2b138094981edd615dbfc4b7787693dbc396</td><td>Statistical Methods For Facial +<br/>Shape-from-shading and Recognition +<br/>Submitted for the degree of Doctor of Philosophy +<br/>Department of Computer Science +<br/>20th February 2007 +</td></tr><tr><td>ad6745dd793073f81abd1f3246ba4102046da022</td><td></td></tr><tr><td>adf62dfa00748381ac21634ae97710bb80fc2922</td><td>ViFaI: A trained video face indexing scheme +<br/>1. Introduction +<br/>With the increasing prominence of inexpensive +<br/>video recording devices (e.g., digital camcorders and +<br/>video recording smartphones), +<br/>the average user’s +<br/>video collection today is increasing rapidly. With this +<br/>development, there arises a natural desire to rapidly +<br/>access a subset of one’s collection of videos. The solu- +<br/>tion to this problem requires an effective video index- +<br/>ing scheme. In particular, we must be able to easily +<br/>process a video to extract such indexes. +<br/>Today, there also exist large sets of labeled (tagged) +<br/>face images. One important example is an individual’s +<br/>Facebook profile. Such a set of of tagged images of +<br/>one’s self, family, friends, and colleagues represents +<br/>an extremely valuable potential training set. +<br/>In this work, we explore how to leverage the afore- +<br/>mentioned training set to solve the video indexing +<br/>problem. +<br/>2. Problem Statement +<br/>Use a labeled (tagged) training set of face images +<br/>to extract relevant indexes from a collection of videos, +<br/>and use these indexes to answer boolean queries of the +<br/>form: “videos with ‘Person 1’ OP1 ‘Person 2’ OP2 ... +<br/>OP(N-1) ‘Person N’ ”, where ‘Person N’ corresponds +<br/>to a training label (tag) and OPN is a boolean operand +<br/>such as AND, OR, NOT, XOR, and so on. +<br/>3. Proposed Scheme +<br/>In this section, we outline our proposed scheme to +<br/>address the problem we postulate in the previous sec- +<br/>tion. We provide further details about the system im- +<br/>plementation in Section 4. +<br/>At a high level, we subdivide the problem into two +<br/>key phases: the first ”off-line” executed once, and the +<br/>second ”on-line” phase instantiated upon each query. +<br/>For the purposes of this work, we define an index as +<br/>follows: <video id, tag, frame #>. +<br/>3.1. The training phase +<br/>We first outline Phase 1 (the training or “off-line” +<br/>phase): +<br/>1. Use the labeled training set plus an additional set +<br/>of ‘other’ faces to compute the Fisher Linear Dis- +<br/>criminant (FLD) [1]. +<br/>2. Project the training data onto the space defined by +<br/>the eigenvectors returned by the FLD, and train +<br/>a classifier (first nearest neighbour, then SVM if +<br/>required) using the training features. +<br/>3. Iterate through each frame of each video, detect- +<br/>ing faces [2], classifying detected results, and add +<br/>an index if the detected face corresponds to one of +<br/>the labeled classes from the previous step. +<br/>3.2. The query phase +<br/>Now, we outline Phase 2 (the query or “on-line” +<br/>phase): +<br/>1. Key the indexes on their video id. +<br/>2. For each video, evaluate the boolean query for the +<br/>set of corresponding indexes. +<br/>3. Keep videos for which the boolean query evalu- +<br/>ates true, and discard those for which it evaluates +<br/>false. +<br/>4. Implementation Details +<br/>We are implementing the project in C++, leverag- +<br/>ing the OpenCV v2.2 framework [4]. In this section, +<br/>we will highlight some of the critical implementation +<br/>details of our proposed system. +</td></tr><tr><td>bba281fe9c309afe4e5cc7d61d7cff1413b29558</td><td>Social Cognitive and Affective Neuroscience, 2017, 984–992 +<br/>doi: 10.1093/scan/nsx030 +<br/>Advance Access Publication Date: 11 April 2017 +<br/>Original article +<br/>An unpleasant emotional state reduces working +<br/>memory capacity: electrophysiological evidence +<br/>1Laboratorio de Neurofisiologia do Comportamento, Departamento de Fisiologia e Farmacologia, Instituto +<br/>Biome´dico, Universidade Federal Fluminense, Niteroi, Brazil, 2MograbiLab, Departamento de Psicologia, +<br/>Pontifıcia Universidade Catolica do Rio de Janeiro, Rio de Janeiro, Brazil, and 3Laboratorio de Engenharia +<br/>Pulmonar, Programa de Engenharia Biome´dica, COPPE, Universidade Federal do Rio de Janeiro, Rio de Janeiro, Brazil +</td></tr><tr><td>bb557f4af797cae9205d5c159f1e2fdfe2d8b096</td><td></td></tr><tr><td>bb06ef67a49849c169781657be0bb717587990e0</td><td>Impact of Temporal Subsampling on Accuracy and +<br/>Performance in Practical Video Classification +<br/>F. Scheidegger∗†, L. Cavigelli∗, M. Schaffner∗, A. C. I. Malossi†, C. Bekas†, L. Benini∗‡ +<br/>∗ETH Zürich, 8092 Zürich, Switzerland +<br/>†IBM Research - Zürich, 8803 Rüschlikon, Switzerland +<br/>‡Università di Bologna, Italy +</td></tr><tr><td>bb22104d2128e323051fb58a6fe1b3d24a9e9a46</td><td>IAJ=JE BH ==OIEI 1 AIIA?A ?= EBH=JE =EO B?KIAI JDA IK>JA +<br/>ABBA?JELAAII B KH =CHEJD +<br/>==OIEI 7IK=O = B=?E= ANFHAIIE ==OIEI IOIJA ?J=EI JDHAA IJ=CAI B=?A =?GKE +<br/>9DAJDAH KIEC *=OAIE= ?=IIEAH " & IKFFHJ LA?JH =?DEA 58 H AKH= +<br/>HACEI E = IECA ?=IIEAH EI = ? IJH=JACO & 0MALAH J = ?= HACEI +</td></tr><tr><td>bbe1332b4d83986542f5db359aee1fd9b9ba9967</td><td></td></tr><tr><td>bb7f2c5d84797742f1d819ea34d1f4b4f8d7c197</td><td>TO APPEAR IN TPAMI +<br/>From Images to 3D Shape Attributes +</td></tr><tr><td>bbf01aa347982592b3e4c9e4f433e05d30e71305</td><td></td></tr><tr><td>bbf1396eb826b3826c5a800975047beabde2f0de</td><td></td></tr><tr><td>bbd1eb87c0686fddb838421050007e934b2d74ab</td><td></td></tr><tr><td>d73d2c9a6cef79052f9236e825058d5d9cdc1321</td><td>2014-ENST-0040 +<br/>EDITE - ED 130 +<br/>Doctorat ParisTech +<br/>T H È S E +<br/>pour obtenir le grade de docteur délivré par +<br/>TELECOM ParisTech +<br/>Spécialité « Signal et Images » +<br/>présentée et soutenue publiquement par +<br/>le 08 juillet 2014 +<br/>Cutting the Visual World into Bigger Slices for Improved Video +<br/>Concept Detection +<br/>Amélioration de la détection des concepts dans les vidéos par de plus grandes tranches du Monde +<br/>Visuel +<br/>Directeur de thèse : Bernard Mérialdo +<br/>Jury +<br/>M. Philippe-Henri Gosselin, Professeur, INRIA +<br/>M. Georges Quénot, Directeur de recherche CNRS, LIG +<br/>M. Georges Linares, Professeur, LIA +<br/>M. François Brémond, Professeur, INRIA +<br/>M. Bernard Mérialdo, Professeur, EURECOM +<br/>Rapporteur +<br/>Rapporteur +<br/>Examinateur +<br/>Examinateur +<br/>Encadrant +<br/>TELECOM ParisTech +<br/>école de l’Institut Télécom - membre de ParisTech +</td></tr><tr><td>d78077a7aa8a302d4a6a09fb9737ab489ae169a6</td><td></td></tr><tr><td>d7312149a6b773d1d97c0c2b847609c07b5255ec</td><td></td></tr><tr><td>d708ce7103a992634b1b4e87612815f03ba3ab24</td><td>FCVID: Fudan-Columbia Video Dataset +<br/>Available at: http://bigvid.fudan.edu.cn/FCVID/ +<br/>1 OVERVIEW +<br/>Recognizing visual contents in unconstrained videos +<br/>has become a very important problem for many ap- +<br/>plications, such as Web video search and recommen- +<br/>dation, smart content-aware advertising, robotics, etc. +<br/>Existing datasets for video content recognition are +<br/>either small or do not have reliable manual labels. +<br/>In this work, we construct and release a new Inter- +<br/>net video dataset called Fudan-Columbia Video Dataset +<br/>(FCVID), containing 91,223 Web videos (total duration +<br/>4,232 hours) annotated manually according to 239 +<br/>categories. We believe that the release of FCVID can +<br/>stimulate innovative research on this challenging and +<br/>important problem. +<br/>2 COLLECTION AND ANNOTATION +<br/>The categories in FCVID cover a wide range of topics +<br/>like social events (e.g., “tailgate party”), procedural +<br/>events (e.g., “making cake”), objects (e.g., “panda”), +<br/>scenes (e.g., “beach”), etc. These categories were de- +<br/>fined very carefully. Specifically, we conducted user +<br/>surveys and used the organization structures on +<br/>YouTube and Vimeo as references, and browsed nu- +<br/>merous videos to identify categories that satisfy the +<br/>following three criteria: (1) utility — high relevance +<br/>in supporting practical application needs; (2) cover- +<br/>age — a good coverage of the contents that people +<br/>record; and (3) feasibility — likely to be automatically +<br/>recognized in the next several years, and a high +<br/>frequency of occurrence that is sufficient for training +<br/>a recognition algorithm. +<br/>This definition effort led to a set of over 250 candi- +<br/>date categories. For each category, in addition to the +<br/>official name used in the public release, we manually +<br/>defined another alternative name. Videos were then +<br/>downloaded from YouTube searches using the official +<br/>and the alternative names as search terms. The pur- +<br/>pose of using the alternative names was to expand the +<br/>candidate video sets. For each search, we downloaded +<br/>1,000 videos, and after removing duplicate videos and +<br/>some extremely long ones (longer than 30 minutes), +<br/>there were around 1,000–1,500 candidate videos for +<br/>each category. +<br/>All the videos were annotated manually to ensure +<br/>a high precision of the FCVID labels. In order to min- +<br/>imize subjectivity, nearly 20 annotators were involved +<br/>in the task, and a master annotator was assigned to +<br/>monitor the entire process and double-check all the +<br/>found positive videos. Some of the videos are multi- +<br/>labeled, and thus filtering the 1,000–1,500 videos for +<br/>each category with focus on just the single category +<br/>label is not adequate. As checking the existence of all +<br/>the 250+ classes for each video is extremely difficult, +<br/>we use the following strategy to narrow down the “la- +<br/>bel search space” for each video. We first grouped the +<br/>categories according to subjective predictions of label +<br/>co-occurrences, e.g., “wedding reception” & “wed- +<br/>ding ceremony”, “waterfall” & “river”, “hiking” & +<br/>“mountain”, and even “dog” & “birthday”. We then +<br/>annotated the videos not only based on the target cat- +<br/>egory label, but also according to the identified related +<br/>labels. This helped produce a fairly complete label +<br/>set for FCVID, but largely reduced the annotation +<br/>workload. After removing the rare categories with +<br/>less than 100 videos after annotation, the final FCVID +<br/>dataset contains 91,223 videos and 239 categories, +<br/>where 183 are events and 56 are objects, scenes, etc. +<br/>Figure 1 shows the number of videos per category. +<br/>“Dog” has the largest number of positive videos +<br/>(1,136), while “making egg tarts” is the most infre- +<br/>quent category containing only 108 samples. The total +<br/>duration of FCVID is 4,232 hours with an average +<br/>video duration of 167 seconds. Figure 2 further gives +<br/>the average video duration of each category. +<br/>The categories are organized using a hierarchy con- +<br/>taining 11 high-level groups, as visualized in Figure 3. +<br/>3 COMPARISON WITH RELATED DATASETS +<br/>We compare FCVID with the following datasets. Most +<br/>of them have been widely adopted in the existing +<br/>works on video categorization. +<br/>KTH and Weizmann: The KTH [1] and the Weiz- +<br/>mann [2] datasets are well-known benchmarks for +<br/>human action recognition. The former contains 600 +<br/>videos of 6 human actions performed by 25 people +<br/>in four scenarios, and the latter consists of 81 videos +<br/>associated with 9 actions performed by 9 actors. +<br/>Hollywood Human Action: The Hollywood +<br/>dataset [3] contains 8 action classes collected from +<br/>32 Hollywood movies with a total of 430 videos. +</td></tr><tr><td>d7b6bbb94ac20f5e75893f140ef7e207db7cd483</td><td>Griffith Research Online +<br/>https://research-repository.griffith.edu.au +<br/>Face Recognition across Pose: A +<br/>Review +<br/>Author +<br/>Zhang, Paul, Gao, Yongsheng +<br/>Published +<br/>2009 +<br/>Journal Title +<br/>Pattern Recognition +<br/>DOI +<br/>https://doi.org/10.1016/j.patcog.2009.04.017 +<br/>Copyright Statement +<br/>Copyright 2009 Elsevier. This is the author-manuscript version of this paper. Reproduced in accordance +<br/>with the copyright policy of the publisher. Please refer to the journal's website for access to the +<br/>definitive, published version. +<br/>Downloaded from +<br/>http://hdl.handle.net/10072/30193 +</td></tr><tr><td>d78373de773c2271a10b89466fe1858c3cab677f</td><td></td></tr><tr><td>d03265ea9200a993af857b473c6bf12a095ca178</td><td>Multiple deep convolutional neural +<br/>networks averaging for face +<br/>alignment +<br/>Zhouping Yin +<br/>Downloaded From: http://electronicimaging.spiedigitallibrary.org/ on 05/28/2015 Terms of Use: http://spiedl.org/terms </td></tr><tr><td>d0eb3fd1b1750242f3bb39ce9ac27fc8cc7c5af0</td><td></td></tr><tr><td>d03baf17dff5177d07d94f05f5791779adf3cd5f</td><td></td></tr><tr><td>d0144d76b8b926d22411d388e7a26506519372eb</td><td>Improving Regression Performance with Distributional Losses +</td></tr><tr><td>d02e27e724f9b9592901ac1f45830341d37140fe</td><td>DA-GAN: Instance-level Image Translation by Deep Attention Generative +<br/>Adversarial Networks +<br/>The State Universtiy of New York at Buffalo +<br/>The State Universtiy of New York at Buffalo +<br/>Microsoft Research +<br/>Microsoft Research +</td></tr><tr><td>d0a21f94de312a0ff31657fd103d6b29db823caa</td><td>Facial Expression Analysis +</td></tr><tr><td>d03e4e938bcbc25aa0feb83d8a0830f9cd3eb3ea</td><td>Face Recognition with Patterns of Oriented +<br/>Edge Magnitudes +<br/>1 Vesalis Sarl, Clermont Ferrand, France +<br/>2 Gipsa-lab, Grenoble INP, France +</td></tr><tr><td>d00787e215bd74d32d80a6c115c4789214da5edb</td><td>Faster and Lighter Online +<br/>Sparse Dictionary Learning +<br/>Project report +</td></tr><tr><td>be8c517406528edc47c4ec0222e2a603950c2762</td><td>Harrigan / The new handbook of methods in nonverbal behaviour research 02-harrigan-chap02 Page Proof page 7 +<br/>17.6.2005 +<br/>5:45pm +<br/>B A S I C R E S E A RC H +<br/>M E T H O D S A N D +<br/>P RO C E D U R E S +</td></tr><tr><td>be48b5dcd10ab834cd68d5b2a24187180e2b408f</td><td>FOR PERSONAL USE ONLY +<br/>Constrained Low-rank Learning Using Least +<br/>Squares Based Regularization +</td></tr><tr><td>be437b53a376085b01ebd0f4c7c6c9e40a4b1a75</td><td>ISSN (Online) 2321 – 2004 +<br/>ISSN (Print) 2321 – 5526 +<br/> INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH IN ELECTRICAL, ELECTRONICS, INSTRUMENTATION AND CONTROL ENGINEERING +<br/> Vol. 4, Issue 5, May 2016 +<br/>IJIREEICE +<br/>Face Recognition and Retrieval Using Cross +<br/>Age Reference Coding +<br/> BE, DSCE, Bangalore1 +<br/>Assistant Professor, DSCE, Bangalore2 +</td></tr><tr><td>bebea83479a8e1988a7da32584e37bfc463d32d4</td><td>Discovery of Latent 3D Keypoints via +<br/>End-to-end Geometric Reasoning +<br/>Google AI +</td></tr><tr><td>bef503cdfe38e7940141f70524ee8df4afd4f954</td><td></td></tr><tr><td>beab10d1bdb0c95b2f880a81a747f6dd17caa9c2</td><td>DeepDeblur: Fast one-step blurry face images restoration +<br/>Tsinghua Unversity +</td></tr><tr><td>b331ca23aed90394c05f06701f90afd550131fe3</td><td>Zhou et al. EURASIP Journal on Image and Video Processing (2018) 2018:49 +<br/>https://doi.org/10.1186/s13640-018-0287-5 +<br/>EURASIP Journal on Image +<br/>and Video Processing +<br/>R ES EAR CH +<br/>Double regularized matrix factorization for +<br/>image classification and clustering +<br/>Open Access +</td></tr><tr><td>b3cb91a08be4117d6efe57251061b62417867de9</td><td>T. Swearingen and A. Ross. "A label propagation approach for predicting missing biographic labels in +<br/>A Label Propagation Approach for +<br/>Predicting Missing Biographic Labels +<br/>in Face-Based Biometric Records +</td></tr><tr><td>b3c60b642a1c64699ed069e3740a0edeabf1922c</td><td>Max-Margin Object Detection +</td></tr><tr><td>b3f7c772acc8bc42291e09f7a2b081024a172564</td><td> www.ijmer.com Vol. 3, Issue. 5, Sep - Oct. 2013 pp-3225-3230 ISSN: 2249-6645 +<br/>International Journal of Modern Engineering Research (IJMER) +<br/>A novel approach for performance parameter estimation of face +<br/>recognition based on clustering, shape and corner detection +<br/><b></b><br/> +</td></tr><tr><td>b32631f456397462b3530757f3a73a2ccc362342</td><td>Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence (IJCAI-17) +<br/>3069 +</td></tr><tr><td>b3afa234996f44852317af382b98f5f557cab25a</td><td></td></tr><tr><td>df90850f1c153bfab691b985bfe536a5544e438b</td><td>FACE TRACKING ALGORITHM ROBUST TO POSE, +<br/>ILLUMINATION AND FACE EXPRESSION CHANGES: A 3D +<br/>PARAMETRIC MODEL APPROACH +<br/><b></b><br/>via Bramante 65 - 26013, Crema (CR), Italy +<br/>Luigi Arnone, Fabrizio Beverina +<br/>STMicroelectronics - Advanced System Technology Group +<br/>via Olivetti 5 - 20041, Agrate Brianza, Italy +<br/>Keywords: +<br/>Face tracking, expression changes, FACS, illumination changes. +</td></tr><tr><td>df8da144a695269e159fb0120bf5355a558f4b02</td><td>International Journal of Computer Applications (0975 – 8887) +<br/>International Conference on Recent Trends in engineering & Technology - 2013(ICRTET'2013) +<br/>Face Recognition using PCA and Eigen Face +<br/>Approach +<br/>ME EXTC [VLSI & Embedded System] +<br/>Sinhgad Academy of Engineering +<br/>EXTC Department +<br/>Pune, India +</td></tr><tr><td>df577a89830be69c1bfb196e925df3055cafc0ed</td><td>Shift: A Zero FLOP, Zero Parameter Alternative to Spatial Convolutions +<br/>UC Berkeley +</td></tr><tr><td>dfabe7ef245ca68185f4fcc96a08602ee1afb3f7</td><td></td></tr><tr><td>df51dfe55912d30fc2f792561e9e0c2b43179089</td><td>Face Hallucination using Linear Models of Coupled +<br/>Sparse Support +<br/>grid and fuse them to suppress the aliasing caused by under- +<br/>sampling [5], [6]. On the other hand, learning based meth- +<br/>ods use coupled dictionaries to learn the mapping relations +<br/>between low- and high- resolution image pairs to synthesize +<br/>high-resolution images from low-resolution images [4], [7]. +<br/>The research community has lately focused on the latter +<br/>category of super-resolution methods, since they can provide +<br/>higher quality images and larger magnification factors. +</td></tr><tr><td>df80fed59ffdf751a20af317f265848fe6bfb9c9</td><td>1666 +<br/>Learning Deep Sharable and Structural +<br/>Detectors for Face Alignment +</td></tr><tr><td>dfa80e52b0489bc2585339ad3351626dee1a8395</td><td>Human Action Forecasting by Learning Task Grammars +</td></tr><tr><td>dfecaedeaf618041a5498cd3f0942c15302e75c3</td><td>Noname manuscript No. +<br/>(will be inserted by the editor) +<br/>A Recursive Framework for Expression Recognition: From +<br/>Web Images to Deep Models to Game Dataset +<br/>Received: date / Accepted: date +</td></tr><tr><td>df5fe0c195eea34ddc8d80efedb25f1b9034d07d</td><td>Robust Modified Active Shape Model for Automatic Facial Landmark +<br/>Annotation of Frontal Faces +</td></tr><tr><td>df674dc0fc813c2a6d539e892bfc74f9a761fbc8</td><td>IOSR Journal of Computer Engineering (IOSR-JCE) +<br/>e-ISSN: 2278-0661, p- ISSN: 2278-8727Volume 10, Issue 6 (May. - Jun. 2013), PP 21-29 +<br/>www.iosrjournals.org +<br/>An Image Mining System for Gender Classification & Age +<br/>Prediction Based on Facial Features +<br/> 1.Ms.Dhanashri Shirkey , 2Prof.Dr.S.R.Gupta, +<br/>M.E(Scholar),Department Computer Science & Engineering, PRMIT & R, Badnera +<br/>Asstt.Prof. Department Computer Science & Engineering, PRMIT & R, Badnera +</td></tr><tr><td>da4170c862d8ae39861aa193667bfdbdf0ecb363</td><td>Multi-task CNN Model for Attribute Prediction +</td></tr><tr><td>da15344a4c10b91d6ee2e9356a48cb3a0eac6a97</td><td></td></tr><tr><td>da5bfddcfe703ca60c930e79d6df302920ab9465</td><td></td></tr><tr><td>dac2103843adc40191e48ee7f35b6d86a02ef019</td><td>854 +<br/>Unsupervised Celebrity Face Naming in Web Videos +</td></tr><tr><td>dae420b776957e6b8cf5fbbacd7bc0ec226b3e2e</td><td>RECOGNIZING EMOTIONS IN SPONTANEOUS FACIAL EXPRESSIONS +<br/>Institut f¨ur Nachrichtentechnik +<br/>Universit¨at Karlsruhe (TH), Germany +</td></tr><tr><td>daba8f0717f3f47c272f018d0a466a205eba6395</td><td></td></tr><tr><td>daefac0610fdeff415c2a3f49b47968d84692e87</td><td>New Orleans, Louisiana, June 1 - 6, 2018. c(cid:13)2018 Association for Computational Linguistics +<br/>Proceedings of NAACL-HLT 2018, pages 1481–1491 +<br/>1481 +</td></tr><tr><td>b49affdff167f5d170da18de3efa6fd6a50262a2</td><td>Author manuscript, published in "Workshop on Faces in 'Real-Life' Images: Detection, Alignment, and Recognition, Marseille : France +<br/>(2008)" +</td></tr><tr><td>b41374f4f31906cf1a73c7adda6c50a78b4eb498</td><td>This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. +<br/>Iterative Gaussianization: From ICA to +<br/>Random Rotations +</td></tr><tr><td>b4d7ca26deb83cec1922a6964c1193e8dd7270e7</td><td></td></tr><tr><td>b4ee64022cc3ccd14c7f9d4935c59b16456067d3</td><td>Unsupervised Cross-Domain Image Generation +</td></tr><tr><td>b40290a694075868e0daef77303f2c4ca1c43269</td><td>第 40 卷 第 4 期 +<br/>2014 年 4 月 +<br/>自 动 化 学 报 +<br/>ACTA AUTOMATICA SINICA +<br/>Vol. 40, No. 4 +<br/>April, 2014 +<br/>融合局部与全局信息的头发形状模型 +<br/>王 楠 1 艾海舟 1 +<br/>摘 要 头发在人体表观中具有重要作用, 然而, 因为缺少有效的形状模型, 头发分割仍然是一个非常具有挑战性的问题. 本 +<br/>文提出了一种基于部件的模型, 它对头发形状以及环境变化更加鲁棒. 该模型将局部与全局信息相结合以描述头发的形状. 局 +<br/>部模型通过一系列算法构建, 包括全局形状词表生成, 词表分类器学习以及参数优化; 而全局模型刻画不同的发型, 采用支持 +<br/>向量机 (Support vector machine, SVM) 来学习, 它为所有潜在的发型配置部件并确定势函数. 在消费者图片上的实验证明 +<br/>了本文算法在头发形状多变和复杂环境等条件下的准确性与有效性. +<br/>关键词 头发形状建模, 部件模型, 部件配置算法, 支持向量机 +<br/>引用格式 王楠, 艾海舟. 融合局部与全局信息的头发形状模型. 自动化学报, 2014, 40(4): 615−623 +<br/>DOI 10.3724/SP.J.1004.2014.00615 +<br/>Combining Local and Global Information for Hair Shape Modeling +<br/>AI Hai-Zhou1 +</td></tr><tr><td>a2359c0f81a7eb032cff1fe45e3b80007facaa2a</td><td>Towards Structured Analysis of Broadcast Badminton Videos +<br/>C.V.Jawahar +<br/>CVIT, KCIS, IIIT Hyderabad +</td></tr><tr><td>a2d9c9ed29bbc2619d5e03320e48b45c15155195</td><td></td></tr><tr><td>a2b54f4d73bdb80854aa78f0c5aca3d8b56b571d</td><td></td></tr><tr><td>a27735e4cbb108db4a52ef9033e3a19f4dc0e5fa</td><td>Intention from Motion +</td></tr><tr><td>a50b4d404576695be7cd4194a064f0602806f3c4</td><td>In Proceedings of BMVC, Edimburgh, UK, September 2006 +<br/>Efficiently estimating facial expression and +<br/>illumination in appearance-based tracking +<br/>†ESCET, U. Rey Juan Carlos +<br/>C/ Tulip´an, s/n +<br/>28933 M´ostoles, Spain +<br/>‡Facultad Inform´atica, UPM +<br/>Campus de Montegancedo s/n +<br/>28660 Boadilla del Monte, Spain +<br/>http://www.dia.fi.upm.es/~pcr +</td></tr><tr><td>a56c1331750bf3ac33ee07004e083310a1e63ddc</td><td>Vol. xx, pp. x +<br/>c(cid:13) xxxx Society for Industrial and Applied Mathematics +<br/>x–x +<br/>Efficient Point-to-Subspace Query in (cid:96)1 with Application to Robust Object +<br/>Instance Recognition +</td></tr><tr><td>a54e0f2983e0b5af6eaafd4d3467b655a3de52f4</td><td>Face Recognition Using Convolution Filters and +<br/>Neural Networks +<br/>Head, Dept. of E&E,PEC +<br/>Sec-12, Chandigarh – 160012 +<br/>Department of CSE & IT, PEC +<br/>Sec-12, Chandigarh – 160012 +<br/>C.P. Singh +<br/>Physics Department, CFSL, +<br/>Sec-36, Chandigarh - 160036 +<br/>a +<br/>of +<br/>to: (a) +<br/>potential method +</td></tr><tr><td>a55efc4a6f273c5895b5e4c5009eabf8e5ed0d6a</td><td>818 +<br/>Continuous Head Movement Estimator for +<br/>Driver Assistance: Issues, Algorithms, +<br/>and On-Road Evaluations +<br/>Mohan Manubhai Trivedi, Fellow, IEEE +</td></tr><tr><td>a5c04f2ad6a1f7c50b6aa5b1b71c36af76af06be</td><td></td></tr><tr><td>a503eb91c0bce3a83bf6f524545888524b29b166</td><td></td></tr><tr><td>a5a44a32a91474f00a3cda671a802e87c899fbb4</td><td>Moments in Time Dataset: one million +<br/>videos for event understanding +</td></tr><tr><td>bd9eb65d9f0df3379ef96e5491533326e9dde315</td><td></td></tr><tr><td>bd07d1f68486052b7e4429dccecdb8deab1924db</td><td></td></tr><tr><td>bd8e2d27987be9e13af2aef378754f89ab20ce10</td><td></td></tr><tr><td>bd2d7c7f0145028e85c102fe52655c2b6c26aeb5</td><td>Attribute-based People Search: Lessons Learnt from a +<br/>Practical Surveillance System +<br/>Rogerio Feris +<br/>IBM Watson +<br/>http://rogerioferis.com +<br/>Russel Bobbitt +<br/>IBM Watson +<br/>Lisa Brown +<br/>IBM Watson +<br/>IBM Watson +</td></tr><tr><td>bdbba95e5abc543981fb557f21e3e6551a563b45</td><td>International Journal of Computational Intelligence and Applications +<br/>Vol. 17, No. 2 (2018) 1850008 (15 pages) +<br/>#.c The Author(s) +<br/>DOI: 10.1142/S1469026818500086 +<br/>Speeding up the Hyperparameter Optimization of Deep +<br/>Convolutional Neural Networks +<br/>Knowledge Technology, Department of Informatics +<br/>Universit€at Hamburg +<br/>Vogt-K€olln-Str. 30, Hamburg 22527, Germany +<br/>Received 15 August 2017 +<br/>Accepted 23 March 2018 +<br/>Published 18 June 2018 +<br/>Most learning algorithms require the practitioner to manually set the values of many hyper- +<br/>parameters before the learning process can begin. However, with modern algorithms, the +<br/>evaluation of a given hyperparameter setting can take a considerable amount of time and the +<br/>search space is often very high-dimensional. We suggest using a lower-dimensional represen- +<br/>tation of the original data to quickly identify promising areas in the hyperparameter space. This +<br/>information can then be used to initialize the optimization algorithm for the original, higher- +<br/>dimensional data. We compare this approach with the standard procedure of optimizing the +<br/>hyperparameters only on the original input. +<br/>We perform experiments with various state-of-the-art hyperparameter optimization algo- +<br/>rithms such as random search, the tree of parzen estimators (TPEs), sequential model-based +<br/>algorithm con¯guration (SMAC), and a genetic algorithm (GA). Our experiments indicate that +<br/>it is possible to speed up the optimization process by using lower-dimensional data repre- +<br/>sentations at the beginning, while increasing the dimensionality of the input later in the opti- +<br/>mization process. This is independent of the underlying optimization procedure, making the +<br/>approach promising for many existing hyperparameter optimization algorithms. +<br/>Keywords: Hyperparameter optimization; hyperparameter importance; convolutional neural +<br/>networks; genetic algorithm; Bayesian optimization. +<br/>1. Introduction +<br/>The performance of many contemporary machine learning algorithms depends cru- +<br/>cially on the speci¯c initialization of hyperparameters such as the general architec- +<br/>ture, the learning rate, regularization parameters, and many others.1,2 Indeed, +<br/>This is an Open Access article published by World Scienti¯c Publishing Company. It is distributed under +<br/>the terms of the Creative Commons Attribution 4.0 (CC-BY) License. Further distribution of this work is +<br/>permitted, provided the original work is properly cited. +<br/>1850008-1 +<br/>Int. J. Comp. Intel. Appl. 2018.17. Downloaded from www.worldscientific.comby WSPC on 07/18/18. Re-use and distribution is strictly not permitted, except for Open Access articles. </td></tr><tr><td>d1dfdc107fa5f2c4820570e369cda10ab1661b87</td><td>Super SloMo: High Quality Estimation of Multiple Intermediate Frames +<br/>for Video Interpolation +<br/>Erik Learned-Miller1 +<br/>1UMass Amherst +<br/>2NVIDIA 3UC Merced +</td></tr><tr><td>d1a43737ca8be02d65684cf64ab2331f66947207</td><td>IJB–S: IARPA Janus Surveillance Video Benchmark (cid:3) +<br/>Kevin O’Connor z +</td></tr><tr><td>d1082eff91e8009bf2ce933ac87649c686205195</td><td>(will be inserted by the editor) +<br/>Pruning of Error Correcting Output Codes by +<br/>Optimization of Accuracy-Diversity Trade off +<br/>S¨ureyya ¨Oz¨o˘g¨ur Aky¨uz · Terry +<br/>Windeatt · Raymond Smith +<br/>Received: date / Accepted: date +</td></tr><tr><td>d69df51cff3d6b9b0625acdcbea27cd2bbf4b9c0</td><td></td></tr><tr><td>d6102a7ddb19a185019fd2112d2f29d9258f6dec</td><td>Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence (IJCAI-17) +<br/>3721 +</td></tr><tr><td>d6bfa9026a563ca109d088bdb0252ccf33b76bc6</td><td>Unsupervised Temporal Segmentation of Facial Behaviour +<br/>Department of Computer Science and Engineering, IIT Kanpur +</td></tr><tr><td>d6fb606e538763282e3942a5fb45c696ba38aee6</td><td></td></tr><tr><td>bc9003ad368cb79d8a8ac2ad025718da5ea36bc4</td><td>Technische Universit¨at M¨unchen +<br/>Bildverstehen und Intelligente Autonome Systeme +<br/>Facial Expression Recognition With A +<br/>Three-Dimensional Face Model +<br/>Vollst¨andiger Abdruck der von der Fakult¨at f¨ur Informatik der Technischen Uni- +<br/>versit¨at M¨unchen zur Erlangung des akademischen Grades eines +<br/>Doktors der Naturwissenschaften +<br/>genehmigten Dissertation. +<br/>Vorsitzender: +<br/>Univ.-Prof. Dr. Johann Schlichter +<br/>Pr¨ufer der Dissertation: 1. Univ.-Prof. Dr. Bernd Radig (i.R.) +<br/>2. Univ.-Prof. Gudrun J. Klinker, Ph.D. +<br/>Die Dissertation wurde am 04.07.2011 bei der Technischen Universit¨at M¨unchen +<br/>eingereicht und durch die Fakult¨at f¨ur Informatik am 02.12.2011 angenommen. +</td></tr><tr><td>bcc346f4a287d96d124e1163e4447bfc47073cd8</td><td></td></tr><tr><td>bcc172a1051be261afacdd5313619881cbe0f676</td><td>978-1-5090-4117-6/17/$31.00 ©2017 IEEE +<br/>2197 +<br/>ICASSP 2017 +</td></tr><tr><td>bcfeac1e5c31d83f1ed92a0783501244dde5a471</td><td></td></tr><tr><td>bc2852fa0a002e683aad3fb0db5523d1190d0ca5</td><td></td></tr><tr><td>bcb99d5150d792001a7d33031a3bd1b77bea706b</td><td></td></tr><tr><td>bc811a66855aae130ca78cd0016fd820db1603ec</td><td>Towards three-dimensional face recognition in the real +<br/>To cite this version: +<br/>HAL Id: tel-00998798 +<br/>https://tel.archives-ouvertes.fr/tel-00998798 +<br/>Submitted on 2 Jun 2014 +<br/>archive for the deposit and dissemination of sci- +<br/>entific research documents, whether they are pub- +<br/>teaching and research institutions in France or +<br/>destin´ee au d´epˆot et `a la diffusion de documents +<br/>recherche fran¸cais ou ´etrangers, des laboratoires +</td></tr><tr><td>bc9af4c2c22a82d2c84ef7c7fcc69073c19b30ab</td><td>MoCoGAN: Decomposing Motion and Content for Video Generation +<br/>Snap Research +<br/>NVIDIA +</td></tr><tr><td>bcac3a870501c5510df80c2a5631f371f2f6f74a</td><td>CVPR +<br/>#1387 +<br/>000 +<br/>001 +<br/>002 +<br/>003 +<br/>004 +<br/>005 +<br/>006 +<br/>007 +<br/>008 +<br/>009 +<br/>010 +<br/>011 +<br/>012 +<br/>013 +<br/>014 +<br/>015 +<br/>016 +<br/>017 +<br/>018 +<br/>019 +<br/>020 +<br/>021 +<br/>022 +<br/>023 +<br/>024 +<br/>025 +<br/>026 +<br/>027 +<br/>028 +<br/>029 +<br/>030 +<br/>031 +<br/>032 +<br/>033 +<br/>034 +<br/>035 +<br/>036 +<br/>037 +<br/>038 +<br/>039 +<br/>040 +<br/>041 +<br/>042 +<br/>043 +<br/>044 +<br/>045 +<br/>046 +<br/>047 +<br/>048 +<br/>049 +<br/>050 +<br/>051 +<br/>052 +<br/>053 +<br/>CVPR 2013 Submission #1387. CONFIDENTIAL REVIEW COPY. DO NOT DISTRIBUTE. +<br/>CVPR +<br/>#1387 +<br/>Structured Face Hallucination +<br/>Anonymous CVPR submission +<br/>Paper ID 1387 +</td></tr><tr><td>aed321909bb87c81121c841b21d31509d6c78f69</td><td></td></tr><tr><td>ae936628e78db4edb8e66853f59433b8cc83594f</td><td></td></tr><tr><td>ae2cf545565c157813798910401e1da5dc8a6199</td><td>Mahkonen et al. EURASIP Journal on Image and Video +<br/>Processing (2018) 2018:61 +<br/>https://doi.org/10.1186/s13640-018-0303-9 +<br/>EURASIP Journal on Image +<br/>and Video Processing +<br/>RESEARCH +<br/>Open Access +<br/>Cascade of Boolean detector +<br/>combinations +</td></tr><tr><td>aebb9649bc38e878baef082b518fa68f5cda23a5</td><td> +</td></tr><tr><td>aeff403079022683b233decda556a6aee3225065</td><td>DeepFace: Face Generation using Deep Learning +</td></tr><tr><td>ae753fd46a744725424690d22d0d00fb05e53350</td><td>000 +<br/>001 +<br/>002 +<br/>003 +<br/>004 +<br/>005 +<br/>006 +<br/>007 +<br/>008 +<br/>009 +<br/>010 +<br/>011 +<br/>012 +<br/>013 +<br/>014 +<br/>015 +<br/>016 +<br/>017 +<br/>018 +<br/>019 +<br/>020 +<br/>021 +<br/>022 +<br/>023 +<br/>024 +<br/>025 +<br/>026 +<br/>027 +<br/>028 +<br/>029 +<br/>030 +<br/>031 +<br/>032 +<br/>033 +<br/>034 +<br/>035 +<br/>036 +<br/>037 +<br/>038 +<br/>039 +<br/>040 +<br/>041 +<br/>042 +<br/>043 +<br/>044 +<br/>Describing Clothing by Semantic Attributes +<br/>Anonymous ECCV submission +<br/>Paper ID 727 +</td></tr><tr><td>ae4e2c81c8a8354c93c4b21442c26773352935dd</td><td></td></tr><tr><td>ae85c822c6aec8b0f67762c625a73a5d08f5060d</td><td>This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication. +<br/>The final version of record is available at http://dx.doi.org/10.1109/TPAMI.2014.2353624 +<br/>IEEE TRANSACTION ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. M, NO. N, MONTH YEAR +<br/>Retrieving Similar Styles to Parse Clothing +</td></tr><tr><td>d861c658db2fd03558f44c265c328b53e492383a</td><td>Automated Face Extraction and Normalization of 3D Mesh Data +</td></tr><tr><td>d83d2fb5403c823287f5889b44c1971f049a1c93</td><td>Motiv Emot +<br/>DOI 10.1007/s11031-013-9353-6 +<br/>O R I G I N A L P A P E R +<br/>Introducing the sick face +<br/>Ó Springer Science+Business Media New York 2013 +</td></tr><tr><td>d8b568392970b68794a55c090c4dd2d7f90909d2</td><td>PDA Face Recognition System +<br/>Using Advanced Correlation +<br/>Filters +<br/>Chee Kiat Ng +<br/>2005 +<br/>Advisor: Prof. Khosla/Reviere +</td></tr><tr><td>d83ae5926b05894fcda0bc89bdc621e4f21272da</td><td>version of the following thesis: +<br/>Frugal Forests: Learning a Dynamic and Cost Sensitive +<br/>Feature Extraction Policy for Anytime Activity Classification +</td></tr><tr><td>d89cfed36ce8ffdb2097c2ba2dac3e2b2501100d</td><td>Robust Face Recognition via Multimodal Deep +<br/>Face Representation +</td></tr><tr><td>ab8f9a6bd8f582501c6b41c0e7179546e21c5e91</td><td>Nonparametric Face Verification Using a Novel +<br/>Face Representation +</td></tr><tr><td>ab58a7db32683aea9281c188c756ddf969b4cdbd</td><td>Efficient Solvers for Sparse Subspace Clustering +</td></tr><tr><td>ab989225a55a2ddcd3b60a99672e78e4373c0df1</td><td>Sample, Computation vs Storage Tradeoffs for +<br/>Classification Using Tensor Subspace Models +</td></tr><tr><td>ab6776f500ed1ab23b7789599f3a6153cdac84f7</td><td>International Journal of Scientific & Engineering Research, Volume 6, Issue 4, April-2015 1212 +<br/>ISSN 2229-5518 +<br/>A Survey on Various Facial Expression +<br/>Techniques +</td></tr><tr><td>ab2b09b65fdc91a711e424524e666fc75aae7a51</td><td>Multi-modal Biomarkers to Discriminate Cognitive State* +<br/>1MIT Lincoln Laboratory, Lexington, Massachusetts, USA +<br/>2USARIEM, 3NSRDEC +<br/>1. Introduction +<br/>Multimodal biomarkers based on behavorial, neurophysiolgical, and cognitive measurements have +<br/>recently obtained increasing popularity in the detection of cognitive stress- and neurological-based +<br/>disorders. Such conditions are significantly and adversely affecting human performance and quality +<br/>of life for a large fraction of the world’s population. Example modalities used in detection of these +<br/>conditions include voice, facial expression, physiology, eye tracking, gait, and EEG analysis. +<br/>Toward the goal of finding simple, noninvasive means to detect, predict and monitor cognitive +<br/>stress and neurological conditions, MIT Lincoln Laboratory is developing biomarkers that satisfy +<br/>three criteria. First, we seek biomarkers that reflect core components of cognitive status such as +<br/>working memory capacity, processing speed, attention, and arousal. Second, and as importantly, we +<br/>seek biomarkers that reflect timing and coordination relations both within components of each +<br/>modality and across different modalities. This is based on the hypothesis that neural coordination +<br/>across different parts of the brain is essential in cognition (Figure 1). An example of timing and +<br/>coordination within a modality is the set of finely timed and synchronized physiological +<br/>components of speech production, while an example of coordination across modalities is the timing +<br/>and synchrony that occurs across speech and facial expression while speaking. Third, we seek +<br/>multimodal biomarkers that contribute in a complementary fashion under various channel and +<br/>background conditions. In this chapter, as an illustration of this biomarker approach we focus on +<br/>cognitive stress and the particular case of detecting different cognitive load levels. We also briefly +<br/>show how similar feature-extraction principles can be applied to a neurological condition through +<br/>the example of major depression disorder (MDD). MDD is one of several neurological disorders +<br/>where multi-modal biomarkers based on principles of timing and coordination are important for +<br/>detection [11]-[22]. In our cognitive load experiments, we use two easily obtained noninvasive +<br/>modalities, voice and face, and show how these two modalities can be fused to produce results on +<br/>par with more invasive, “gold-standard” EEG measurements. Vocal and facial biomarkers will also +<br/>be used in our MDD case study. In both application areas we focus on timing and coordination +<br/>relations within the components of each modality. +<br/>* Distribution A: public release.This work is sponsored by the Assistant Secretary of Defense for Research & Engineering under Air Force contract +<br/>#FA8721-05-C-0002. Opinions,interpretations, conclusions, and recommendations are those of the authors and are not necessarily endorsed by the United States +<br/>Government. +</td></tr><tr><td>ab87dfccb1818bdf0b41d732da1f9335b43b74ae</td><td>SUBMITTED TO IEEE TRANSACTIONS ON SIGNAL PROCESSING +<br/>Structured Dictionary Learning for Classification +</td></tr><tr><td>ab1dfcd96654af0bf6e805ffa2de0f55a73c025d</td><td></td></tr><tr><td>abeda55a7be0bbe25a25139fb9a3d823215d7536</td><td>UNIVERSITATPOLITÈCNICADECATALUNYAProgramadeDoctorat:AUTOMÀTICA,ROBÒTICAIVISIÓTesiDoctoralUnderstandingHuman-CentricImages:FromGeometrytoFashionEdgarSimoSerraDirectors:FrancescMorenoNoguerCarmeTorrasMay2015 </td></tr><tr><td>ab1900b5d7cf3317d17193e9327d57b97e24d2fc</td><td></td></tr><tr><td>ab8fb278db4405f7db08fa59404d9dd22d38bc83</td><td>UNIVERSITÉ DE GENÈVE +<br/>Département d'Informatique +<br/>FACULTÉ DES SCIENCES +<br/>Implicit and Automated Emotional +<br/>Tagging of Videos +<br/>THÈSE +<br/>présenté à la Faculté des sciences de l'Université de Genève +<br/>pour obtenir le grade de Docteur ès sciences, mention informatique +<br/>par +<br/>de +<br/>Téhéran (IRAN) +<br/>Thèse No 4368 +<br/>GENÈVE +<br/>Repro-Mail - Université de Genève +<br/>2011 +</td></tr><tr><td>e5737ffc4e74374b0c799b65afdbf0304ff344cb</td><td></td></tr><tr><td>e5823a9d3e5e33e119576a34cb8aed497af20eea</td><td>DocFace+: ID Document to Selfie* Matching +</td></tr><tr><td>e5dfd17dbfc9647ccc7323a5d62f65721b318ba9</td><td></td></tr><tr><td>e56c4c41bfa5ec2d86c7c9dd631a9a69cdc05e69</td><td>Human Activity Recognition Based on Wearable +<br/>Sensor Data: A Standardization of the +<br/>State-of-the-Art +<br/>Smart Surveillance Interest Group, Computer Science Department +<br/>Universidade Federal de Minas Gerais, Brazil +</td></tr><tr><td>e27c92255d7ccd1860b5fb71c5b1277c1648ed1e</td><td></td></tr><tr><td>e200c3f2849d56e08056484f3b6183aa43c0f13a</td><td></td></tr><tr><td>f437b3884a9e5fab66740ca2a6f1f3a5724385ea</td><td>Human Identification Technical Challenges +<br/>DARPA +<br/>3701 N. Fairfax Dr +<br/>Arlington, VA 22203 +</td></tr><tr><td>f442a2f2749f921849e22f37e0480ac04a3c3fec</td><td></td></tr><tr><td>f4f6fc473effb063b7a29aa221c65f64a791d7f4</td><td>Downloaded From: https://www.spiedigitallibrary.org/journals/Journal-of-Electronic-Imaging on 4/20/2018 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use +<br/>FacialexpressionrecognitioninthewildbasedonmultimodaltexturefeaturesBoSunLiandongLiGuoyanZhouJunHeBoSun,LiandongLi,GuoyanZhou,JunHe,“Facialexpressionrecognitioninthewildbasedonmultimodaltexturefeatures,”J.Electron.Imaging25(6),061407(2016),doi:10.1117/1.JEI.25.6.061407. </td></tr><tr><td>f4c01fc79c7ead67899f6fe7b79dd1ad249f71b0</td><td></td></tr><tr><td>f4373f5631329f77d85182ec2df6730cbd4686a9</td><td>Soft Computing manuscript No. +<br/>(will be inserted by the editor) +<br/>Recognizing Gender from Human Facial Regions using +<br/>Genetic Algorithm +<br/>Received: date / Accepted: date +</td></tr><tr><td>f47404424270f6a20ba1ba8c2211adfba032f405</td><td>International Journal of Emerging Technology and Advanced Engineering +<br/>Website: www.ijetae.com (ISSN 2250-2459, Volume 2, Issue 5, May 2012) +<br/>Identification of Face Age range Group using Neural +<br/>Network +</td></tr><tr><td>f3fcaae2ea3e998395a1443c87544f203890ae15</td><td></td></tr><tr><td>f3d9e347eadcf0d21cb0e92710bc906b22f2b3e7</td><td>NosePose: a competitive, landmark-free +<br/>methodology for head pose estimation in the wild +<br/>IMAGO Research Group - Universidade Federal do Paran´a +</td></tr><tr><td>f355e54ca94a2d8bbc598e06e414a876eb62ef99</td><td></td></tr><tr><td>f3ea181507db292b762aa798da30bc307be95344</td><td>Covariance Pooling for Facial Expression Recognition +<br/>†Computer Vision Lab, ETH Zurich, Switzerland +<br/>‡VISICS, KU Leuven, Belgium +</td></tr><tr><td>f3cf10c84c4665a0b28734f5233d423a65ef1f23</td><td>Title +<br/>Temporal Exemplar-based Bayesian Networks for facial +<br/>expression recognition +<br/>Author(s) +<br/>Shang, L; Chan, KP +<br/>Citation +<br/>Proceedings - 7Th International Conference On Machine +<br/>Learning And Applications, Icmla 2008, 2008, p. 16-22 +<br/>Issued Date +<br/>2008 +<br/>URL +<br/>http://hdl.handle.net/10722/61208 +<br/>Rights +<br/>This work is licensed under a Creative Commons Attribution- +<br/>NonCommercial-NoDerivatives 4.0 International License.; +<br/>International Conference on Machine Learning and Applications +<br/>Proceedings. Copyright © IEEE.; ©2008 IEEE. Personal use of +<br/>this material is permitted. However, permission to +<br/>reprint/republish this material for advertising or promotional +<br/>purposes or for creating new collective works for resale or +<br/>redistribution to servers or lists, or to reuse any copyrighted +<br/>component of this work in other works must be obtained from +<br/>the IEEE. +</td></tr><tr><td>f3b7938de5f178e25a3cf477107c76286c0ad691</td><td>JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, MARCH 2017 +<br/>Object Detection with Deep Learning: A Review +</td></tr><tr><td>ebedc841a2c1b3a9ab7357de833101648281ff0e</td><td></td></tr><tr><td>eb526174fa071345ff7b1fad1fad240cd943a6d7</td><td>Deeply Vulnerable – A Study of the Robustness of Face Recognition to +<br/>Presentation Attacks +</td></tr><tr><td>eb566490cd1aa9338831de8161c6659984e923fd</td><td>From Lifestyle Vlogs to Everyday Interactions +<br/>EECS Department, UC Berkeley +</td></tr><tr><td>eb9312458f84a366e98bd0a2265747aaed40b1a6</td><td>1-4244-1437-7/07/$20.00 ©2007 IEEE +<br/>IV - 473 +<br/>ICIP 2007 +</td></tr><tr><td>eb716dd3dbd0f04e6d89f1703b9975cad62ffb09</td><td>Copyright +<br/>by +<br/>2012 +</td></tr><tr><td>ebabd1f7bc0274fec88a3dabaf115d3e226f198f</td><td>Driver drowsiness detection system based on feature +<br/>representation learning using various deep networks +<br/>School of Electrical Engineering, KAIST, +<br/>Guseong-dong, Yuseong-gu, Dajeon, Rep. of Korea +</td></tr><tr><td>ebb9d53668205c5797045ba130df18842e3eadef</td><td></td></tr><tr><td>eb48a58b873295d719827e746d51b110f5716d6c</td><td>Face Alignment Using K-cluster Regression Forests +<br/>With Weighted Splitting +</td></tr><tr><td>c7e4c7be0d37013de07b6d829a3bf73e1b95ad4e</td><td>The International Journal of Multimedia & Its Applications (IJMA) Vol.5, No.5, October 2013 +<br/>DYNEMO: A VIDEO DATABASE OF NATURAL FACIAL +<br/>EXPRESSIONS OF EMOTIONS +<br/>1LIP, Univ. Grenoble Alpes, BP 47 - 38040 Grenoble Cedex 9, France +<br/>2LIG, Univ. Grenoble Alpes, BP 53 - 38041 Grenoble Cedex 9, France +</td></tr><tr><td>c7c5f0fe1fcaf3787c7f78f7dc62f3497dcfdf3c</td><td>THE IMPACT OF PRODUCT PHOTO ON ONLINE CONSUMER +<br/>PURCHASE INTENTION: AN IMAGE-PROCESSING ENABLED +<br/>EMPIRICAL STUDY +</td></tr><tr><td>c758b9c82b603904ba8806e6193c5fefa57e9613</td><td>Heterogeneous Face Recognition with CNNs +<br/>INRIA Grenoble, Laboratoire Jean Kuntzmann +</td></tr><tr><td>c7c8d150ece08b12e3abdb6224000c07a6ce7d47</td><td>DeMeshNet: Blind Face Inpainting for Deep MeshFace Verification +<br/>National Laboratory of Pattern Recognition, CASIA +<br/>Center for Research on Intelligent Perception and Computing, CASIA +</td></tr><tr><td>c038beaa228aeec174e5bd52460f0de75e9cccbe</td><td>Temporal Segment Networks for Action +<br/>Recognition in Videos +</td></tr><tr><td>c043f8924717a3023a869777d4c9bee33e607fb5</td><td>Emotion Separation Is Completed Early and It Depends +<br/>on Visual Field Presentation +<br/><b>Lab for Human Brain Dynamics, RIKEN Brain Science Institute, Wakoshi, Saitama, Japan, 2 Lab for Human Brain Dynamics, AAI Scientific Cultural Services Ltd., Nicosia</b><br/>Cyprus +</td></tr><tr><td>c05a7c72e679745deab9c9d7d481f7b5b9b36bdd</td><td>NPS-CS-11-005 +<br/> +<br/> +<br/>NAVAL +<br/>POSTGRADUATE +<br/>SCHOOL +<br/>MONTEREY, CALIFORNIA +<br/>by +<br/>BIOMETRIC CHALLENGES FOR FUTURE DEPLOYMENTS: +<br/>A STUDY OF THE IMPACT OF GEOGRAPHY, CLIMATE, CULTURE, +<br/> AND SOCIAL CONDITIONS ON THE EFFECTIVE +<br/>COLLECTION OF BIOMETRICS +<br/>April 2011 +<br/>Approved for public release; distribution is unlimited +</td></tr><tr><td>c02847a04a99a5a6e784ab580907278ee3c12653</td><td>Fine Grained Video Classification for +<br/>Endangered Bird Species Protection +<br/>Non-Thesis MS Final Report +<br/>1. Introduction +<br/>1.1 Background +<br/>This project is about detecting eagles in videos. Eagles are endangered species at the brim of +<br/>extinction since 1980s. With the bans of harmful pesticides, the number of eagles keep increasing. +<br/>However, recent studies on golden eagles’ activities in the vicinity of wind turbines have shown +<br/>significant number of turbine blade collisions with eagles as the major cause of eagles’ mortality. [1] +<br/>This project is a part of a larger research project to build an eagle detection and deterrent system +<br/>on wind turbine toward reducing eagles’ mortality. [2] The critical component of this study is a +<br/>computer vision system for eagle detection in videos. The key requirement are that the system should +<br/>work in real time and detect eagles at a far distance from the camera (i.e. in low resolution). +<br/>There are three different bird species in my dataset - falcon, eagle and seagull. The reason for +<br/>involving only these three species is based on the real world situation. Wind turbines are always +<br/>installed near coast and mountain hill where falcons and seagulls will be the majority. So my model +<br/>will classify the minority eagles out of other bird species during the immigration season and protecting +<br/>them by using the deterrent system. +<br/>1.2 Brief Approach +<br/>Our approach represents a unified deep-learning architecture for eagle detection. Given videos, +<br/>our goal is to detect eagle species at far distance from the camera, using both appearance and bird +<br/>motion cues, so as to meet the recall-precision rates set by the user. Detecting eagle is a challenging +<br/>task because of the following reasons. Frist, an eagle flies fast and high in the sky which means that +<br/>we need a lens with wide angle such that captures their movement. However, a camera with wide +<br/>angle produces a low resolution and low quality video and the detailed appearance of bird is +<br/>compromised. Second, current neural network typically take as input low resolution images. This is +<br/>because a higher resolution image will require larger filters and deeper networks which is turn hard to +<br/>train [3]. So it is not clear whether the low resolution will cause challenge for fine-grained +<br/>classification task. Last but not the least, there is not a large training database like PASCAL, MNIST +</td></tr><tr><td>c0c8d720658374cc1ffd6116554a615e846c74b5</td><td>JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 +<br/>Modeling Multimodal Clues in a Hybrid Deep +<br/>Learning Framework for Video Classification +</td></tr><tr><td>c0d5c3aab87d6e8dd3241db1d931470c15b9e39d</td><td></td></tr><tr><td>eee8a37a12506ff5df72c402ccc3d59216321346</td><td>Uredniki: +<br/>dr. Tomaž Erjavec +<br/>Odsek za tehnologije znanja +<br/>Institut »Jožef Stefan«, Ljubljana +<br/>dr. Jerneja Žganec Gros +<br/>Alpineon d.o.o, Ljubljana +<br/>Založnik: Institut »Jožef Stefan«, Ljubljana +<br/>Tisk: Birografika BORI d.o.o. +<br/>Priprava zbornika: Mitja Lasič +<br/>Oblikovanje naslovnice: dr. Damjan Demšar +<br/>Tiskano iz predloga avtorjev +<br/>Naklada: 50 +<br/>Ljubljana, oktober 2008 +<br/>Konferenco IS 2008 sofinancirata +<br/>Ministrstvo za visoko šolstvo, znanost in tehnologijo +<br/>Institut »Jožef Stefan« +<br/>ISSN 1581-9973 +<br/>CIP - Kataložni zapis o publikaciji +<br/>Narodna in univerzitetna knjižnica, Ljubljana +<br/>004.934(082) +<br/>81'25:004.6(082) +<br/>004.8(063) +<br/>oktober 2008, Ljubljana, Slovenia : zbornik 11. mednarodne +<br/>Proceedings of the Sixth Language Technologies Conference, October +<br/>16th-17th, 2008 : proceedings of the 11th International +<br/>Multiconference Information Society - IS 2008, volume C / uredila, +<br/>edited by Tomaž Erjavec, Jerneja Žganec Gros. - Ljubljana : +<br/>1581-9973) +<br/>ISBN 978-961-264-006-4 +<br/>družba 4. Information society 5. Erjavec, Tomaž, 1960- 6. +<br/>Ljubljana) +<br/>241520896 +</td></tr><tr><td>ee18e29a2b998eddb7f6663bb07891bfc7262248</td><td>1119 +<br/>Local Linear Discriminant Analysis Framework +<br/>Using Sample Neighbors +</td></tr><tr><td>ee461d060da58d6053d2f4988b54eff8655ecede</td><td></td></tr><tr><td>eefb8768f60c17d76fe156b55b8a00555eb40f4d</td><td>Subspace Scores for Feature Selection in Computer Vision +</td></tr><tr><td>eed1dd2a5959647896e73d129272cb7c3a2e145c</td><td></td></tr><tr><td>ee92d36d72075048a7c8b2af5cc1720c7bace6dd</td><td>FACE RECOGNITION USING MIXTURES OF PRINCIPAL COMPONENTS +<br/>Video and Display Processing +<br/>Philips Research USA +<br/>Briarcliff Manor, NY 10510 +</td></tr><tr><td>eedfb384a5e42511013b33104f4cd3149432bd9e</td><td>Multimodal Probabilistic Person +<br/>Tracking and Identification +<br/>in Smart Spaces +<br/>zur Erlangung des akademischen Grades eines +<br/>Doktors der Ingenieurwissenschaften +<br/>der Fakultät für Informatik +<br/>der Universität Fridericiana zu Karlsruhe (TH) +<br/>genehmigte +<br/>Dissertation +<br/>von +<br/>aus Karlsruhe +<br/>Tag der mündlichen Prüfung: 20.11.2009 +<br/>Erster Gutachter: +<br/>Zweiter Gutachter: +<br/>Prof. Dr. A. Waibel +<br/>Prof. Dr. R. Stiefelhagen +</td></tr><tr><td>c9424d64b12a4abe0af201e7b641409e182babab</td><td>Article +<br/>Which, When, and How: Hierarchical Clustering with +<br/>Human–Machine Cooperation +<br/>Academic Editor: Tom Burr +<br/>Received: 3 November 2016; Accepted: 14 December 2016; Published: 21 December 2016 +</td></tr><tr><td>c903af0d69edacf8d1bff3bfd85b9470f6c4c243</td><td></td></tr><tr><td>fc1e37fb16006b62848def92a51434fc74a2431a</td><td>DRAFT +<br/>A Comprehensive Analysis of Deep Regression +</td></tr><tr><td>fc516a492cf09aaf1d319c8ff112c77cfb55a0e5</td><td></td></tr><tr><td>fcd3d69b418d56ae6800a421c8b89ef363418665</td><td>Effects of Aging over Facial Feature Analysis and Face +<br/>Recognition +<br/>Bogaziçi Un. Electronics Eng. Dept. March 2010 +</td></tr><tr><td>fcd77f3ca6b40aad6edbd1dab9681d201f85f365</td><td>c(cid:13)Copyright 2014 +</td></tr><tr><td>fcf8bb1bf2b7e3f71fb337ca3fcf3d9cf18daa46</td><td>MANUSCRIPT SUBMITTED TO IEEE TRANS. PATTERN ANAL. MACH. INTELL., JULY 2010 +<br/>Feature Selection via Sparse Approximation for +<br/>Face Recognition +</td></tr><tr><td>fcbf808bdf140442cddf0710defb2766c2d25c30</td><td>IJCV manuscript No. +<br/>(will be inserted by the editor) +<br/>Unsupervised Semantic Action Discovery from Video +<br/>Collections +<br/>Received: date / Accepted: date +</td></tr><tr><td>fd4ac1da699885f71970588f84316589b7d8317b</td><td>JOURNAL OF LATEX CLASS FILES, VOL. 6, NO. 1, JANUARY 2007 +<br/>Supervised Descent Method +<br/>for Solving Nonlinear Least Squares +<br/>Problems in Computer Vision +</td></tr><tr><td>fdf533eeb1306ba418b09210387833bdf27bb756</td><td>951 +</td></tr><tr><td>fdda5852f2cffc871fd40b0cb1aa14cea54cd7e3</td><td>Im2Flow: Motion Hallucination from Static Images for Action Recognition +<br/>UT Austin +<br/>UT Austin +<br/>UT Austin +</td></tr><tr><td>fdfaf46910012c7cdf72bba12e802a318b5bef5a</td><td>Computerized Face Recognition in Renaissance +<br/>Portrait Art +</td></tr><tr><td>fd15e397629e0241642329fc8ee0b8cd6c6ac807</td><td>Semi-Supervised Clustering with Neural Networks +<br/>IIIT-Delhi, India +</td></tr><tr><td>fdca08416bdadda91ae977db7d503e8610dd744f</td><td> +<br/>ICT-2009.7.1 +<br/>KSERA Project +<br/>2010-248085 +<br/>Deliverable D3.1 +<br/>Deliverable D3.1 +<br/>Human Robot Interaction +<br/>Human Robot Interaction +<br/>18 October 2010 +<br/>Public Document +<br/>The KSERA project (http://www.ksera +<br/>KSERA project (http://www.ksera-project.eu) has received funding from the European Commission +<br/>project.eu) has received funding from the European Commission +<br/>under the 7th Framework Programme (FP7) for Research and Technological Development under grant +<br/>under the 7th Framework Programme (FP7) for Research and Technological Development under grant +<br/>under the 7th Framework Programme (FP7) for Research and Technological Development under grant +<br/>agreement n°2010-248085. +</td></tr><tr><td>fdaf65b314faee97220162980e76dbc8f32db9d6</td><td>Accepted Manuscript +<br/>Face recognition using both visible light image and near-infrared image and a deep +<br/>network +<br/>PII: +<br/>DOI: +<br/>Reference: +<br/>S2468-2322(17)30014-8 +<br/>10.1016/j.trit.2017.03.001 +<br/>TRIT 41 +<br/>To appear in: +<br/>CAAI Transactions on Intelligence Technology +<br/>Received Date: 30 January 2017 +<br/>Accepted Date: 28 March 2017 +<br/>Please cite this article as: K. Guo, S. Wu, Y. Xu, Face recognition using both visible light image and +<br/>near-infrared image and a deep network, CAAI Transactions on Intelligence Technology (2017), doi: +<br/>10.1016/j.trit.2017.03.001. +<br/>This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to +<br/>our customers we are providing this early version of the manuscript. The manuscript will undergo +<br/>copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please +<br/>note that during the production process errors may be discovered which could affect the content, and all +<br/>legal disclaimers that apply to the journal pertain. +</td></tr><tr><td>f2e9494d0dca9fb6b274107032781d435a508de6</td><td></td></tr><tr><td>f2c568fe945e5743635c13fe5535af157b1903d1</td><td></td></tr><tr><td>f26097a1a479fb6f32b27a93f8f32609cfe30fdc</td><td></td></tr><tr><td>f231046d5f5d87e2ca5fae88f41e8d74964e8f4f</td><td>We are IntechOpen, +<br/>the first native scientific +<br/>publisher of Open Access books +<br/>3,350 +<br/>108,000 +<br/>1.7 M +<br/>Open access books available +<br/>International authors and editors +<br/>Downloads +<br/>Our authors are among the +<br/>151 +<br/>Countries delivered to +<br/>TOP 1% +<br/>12.2% +<br/>most cited scientists +<br/>Contributors from top 500 universities +<br/>Selection of our books indexed in the Book Citation Index +<br/>in Web of Science™ Core Collection (BKCI) +<br/>Interested in publishing with us? +<br/>Numbers displayed above are based on latest data collected. +<br/>For more information visit www.intechopen.com +</td></tr><tr><td>f214bcc6ecc3309e2efefdc21062441328ff6081</td><td></td></tr><tr><td>f5770dd225501ff3764f9023f19a76fad28127d4</td><td>Real Time Online Facial Expression Transfer +<br/>with Single Video Camera +</td></tr><tr><td>f519723238701849f1160d5a9cedebd31017da89</td><td>Impact of multi-focused images on recognition of soft biometric traits +<br/>aEURECOM, Campus SophiaTech, 450 Route des Chappes, CS 50193 - 06904 Biot Sophia +<br/> +<br/>Antipolis cedex, FRANCE +</td></tr><tr><td>f558af209dd4c48e4b2f551b01065a6435c3ef33</td><td>International Journal of Emerging Technology in Computer Science & Electronics (IJETCSE) +<br/>ISSN: 0976-1353 Volume 23 Issue 1 –JUNE 2016. +<br/>AN ENHANCED ATTRIBUTE +<br/>RERANKING DESIGN FOR WEB IMAGE +<br/>SEARCH +<br/>#Student,Cse, CIET, Lam,Guntur, India +<br/>* Assistant Professort,Cse, CIET, Lam,Guntur , India +</td></tr><tr><td>e393a038d520a073b9835df7a3ff104ad610c552</td><td>Automatic temporal segment +<br/>detection via bilateral long short- +<br/>term memory recurrent neural +<br/>networks +<br/>detection via bilateral long short-term memory recurrent neural networks,” J. +<br/>Electron. Imaging 26(2), 020501 (2017), doi: 10.1117/1.JEI.26.2.020501. +<br/>Downloaded From: http://electronicimaging.spiedigitallibrary.org/ on 03/03/2017 Terms of Use: http://spiedigitallibrary.org/ss/termsofuse.aspx </td></tr><tr><td>e3657ab4129a7570230ff25ae7fbaccb4ba9950c</td><td></td></tr><tr><td>e315959d6e806c8fbfc91f072c322fb26ce0862b</td><td>An Efficient Face Recognition System Based on Sub-Window +<br/>International Journal of Soft Computing and Engineering (IJSCE) +<br/>ISSN: 2231-2307, Volume-1, Issue-6, January 2012 +<br/>Extraction Algorithm +</td></tr><tr><td>e3c011d08d04c934197b2a4804c90be55e21d572</td><td>How to Train Triplet Networks with 100K Identities? +<br/>Orion Star +<br/>Beijing, China +<br/>Orion Star +<br/>Beijing, China +<br/>Orion Star +<br/>Beijing, China +</td></tr><tr><td>e39a0834122e08ba28e7b411db896d0fdbbad9ba</td><td>1368 +<br/>Maximum Likelihood Estimation of Depth Maps +<br/>Using Photometric Stereo +</td></tr><tr><td>e3917d6935586b90baae18d938295e5b089b5c62</td><td>152 +<br/>Face Localization and Authentication +<br/>Using Color and Depth Images +</td></tr><tr><td>cfa572cd6ba8dfc2ee8ac3cc7be19b3abff1a8a2</td><td></td></tr><tr><td>cfffae38fe34e29d47e6deccfd259788176dc213</td><td>TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. X, NO. X, DECEMBER 2012 +<br/>Matrix Completion for Weakly-supervised +<br/>Multi-label Image Classification +</td></tr><tr><td>cfd4004054399f3a5f536df71f9b9987f060f434</td><td>IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. ??, NO. ??, ?? 20?? +<br/>Person Recognition in Personal Photo Collections +</td></tr><tr><td>cfb8bc66502fb5f941ecdb22aec1fdbfdb73adce</td><td></td></tr><tr><td>cf875336d5a196ce0981e2e2ae9602580f3f6243</td><td>7 What 1 +<br/>Rosalind W. Picard +<br/>It Mean for a Computer to "Have" Emotions? +<br/>There is a lot of talk about giving machines emotions, some of +<br/>it fluff. Recently at a large technical meeting, a researcher stood up +<br/>and talked of how a Bamey stuffed animal [the purple dinosaur for +<br/>kids) "has emotions." He did not define what he meant by this, but +<br/>after repeating it several times, it became apparent that children +<br/>attributed emotions to Barney, and that Barney had deliberately +<br/>expressive behaviors that would encourage the kids to think. Bar- +<br/>ney had emotions. But kids have attributed emotions to dolls and +<br/>stuffed animals for as long a s we know; and most of my technical +<br/>colleagues would agree that such toys have never had and still do +<br/>not have emotions. What is different now that prompts a researcher +<br/>to make such a claim? Is the computational plush an example of a +<br/>computer that really does have emotions? +<br/>If not Barney, then what would be an example of a computa- +<br/>tional system that has emotions? I am not a philosopher, and this +<br/>paper will not be a discussion of the meaning of this question in +<br/>any philosophical sense. However, as an engineer I am interested +<br/>in what capabilities I would require a machine to have before I +<br/>would say that it "has emotions," if that is even possible. +<br/>Theorists still grappl~ with the problem of defining emotion, +<br/>after many decades of discussion, and no clean definition looks +<br/>likely to emerge. Even without a precise definition, one can still +<br/>begin to say concrete things about certain components of emotion, +<br/>at least based on what is known about human and animal emo- +<br/>tions. Of course, much is still u d a o w n about human emotions, so +<br/>we are nowhere near being able to model them, much less dupli- +<br/>cate all their functions in machines.'~lso, all scientific findings are +<br/>subject to revision-history has certainly taught us humility, that +<br/>what scientists believed to be true at one point has often been +<br/>changed at a later date. +<br/>I wish to begin by mentioning four motivations for giving +<br/>machines certain emotional abilities (and there are more). One goal +<br/>is to build robots and synthetic characters that can emulate living +<br/>humans and animals-for example, to build a humanoid robot. A +<br/>I +</td></tr><tr><td>cf54a133c89f730adc5ea12c3ac646971120781c</td><td></td></tr><tr><td>cfbb2d32586b58f5681e459afd236380acd86e28</td><td>Improving Alignment of Faces for Recognition +<br/>Christopher J. Pal +<br/>D´epartement de g´enie informatique et g´enie logiciel +<br/>´Ecole Polytechnique de Montr´eal, +<br/>D´epartement de g´enie informatique et g´enie logiciel +<br/>´Ecole Polytechnique de Montr´eal, +<br/>Qu´ebec, Canada +<br/>Qu´ebec, Canada +</td></tr><tr><td>cfa92e17809e8d20ebc73b4e531a1b106d02b38c</td><td>Advances in Data Analysis and Classification manuscript No. +<br/>(will be inserted by the editor) +<br/>Parametric Classification with Soft Labels using the +<br/>Evidential EM Algorithm +<br/>Linear Discriminant Analysis vs. Logistic Regression +<br/>Received: date / Accepted: date +</td></tr><tr><td>cfdc632adcb799dba14af6a8339ca761725abf0a</td><td>Probabilistic Formulations of Regression with Mixed +<br/>Guidance +</td></tr><tr><td>cfc30ce53bfc204b8764ebb764a029a8d0ad01f4</td><td>Regularizing Deep Neural Networks by Noise: +<br/>Its Interpretation and Optimization +<br/>Dept. of Computer Science and Engineering, POSTECH, Korea +</td></tr><tr><td>cf86616b5a35d5ee777585196736dfafbb9853b5</td><td>This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. +<br/>Learning Multiscale Active Facial Patches for +<br/>Expression Analysis +</td></tr><tr><td>cad52d74c1a21043f851ae14c924ac689e197d1f</td><td>From Ego to Nos-vision: +<br/>Detecting Social Relationships in First-Person Views +<br/>Universit`a degli Studi di Modena e Reggio Emilia +<br/>Via Vignolese 905, 41125 Modena - Italy +</td></tr><tr><td>cac8bb0e393474b9fb3b810c61efdbc2e2c25c29</td><td></td></tr><tr><td>cad24ba99c7b6834faf6f5be820dd65f1a755b29</td><td>Understanding hand-object +<br/>manipulation by modeling the +<br/>contextual relationship between actions, +<br/>grasp types and object attributes +<br/>Journal Title +<br/>XX(X):1–14 +<br/>c(cid:13)The Author(s) 2016 +<br/>Reprints and permission: +<br/>sagepub.co.uk/journalsPermissions.nav +<br/>DOI: 10.1177/ToBeAssigned +<br/>www.sagepub.com/ +</td></tr><tr><td>cadba72aa3e95d6dcf0acac828401ddda7ed8924</td><td>THÈSE PRÉSENTÉE À LA FACULTÉ DES SCIENCES +<br/>POUR L’OBTENTION DU GRADE DE DOCTEUR ÈS SCIENCES +<br/>Algorithms and VLSI Architectures +<br/>for Low-Power Mobile Face Verification +<br/>par +<br/>Acceptée sur proposition du jury: +<br/>Prof. F. Pellandini, directeur de thèse +<br/>PD Dr. M. Ansorge, co-directeur de thèse +<br/>Prof. P.-A. Farine, rapporteur +<br/>Dr. C. Piguet, rapporteur +<br/>Soutenue le 2 juin 2005 +<br/>INSTITUT DE MICROTECHNIQUE +<br/>UNIVERSITÉ DE NEUCHÂTEL +<br/>2006 +</td></tr><tr><td>ca606186715e84d270fc9052af8500fe23befbda</td><td>Using Subclass Discriminant Analysis, Fuzzy Integral and Symlet Decomposition for +<br/>Face Recognition +<br/>Department of Electrical Engineering, +<br/>Iran Univ. of Science and Technology, +<br/>Narmak, Tehran, Iran +<br/>Department of Electrical Engineering, +<br/>Iran Univ. of Science and Technology, +<br/>Department of Electrical Engineering, +<br/>Iran Univ. of Science and Technology, +<br/>Narmak, Tehran, Iran +<br/>Narmak, Tehran, Iran +</td></tr><tr><td>e465f596d73f3d2523dbf8334d29eb93a35f6da0</td><td></td></tr><tr><td>e4aeaf1af68a40907fda752559e45dc7afc2de67</td><td></td></tr><tr><td>e4c3d5d43cb62ac5b57d74d55925bdf76205e306</td><td></td></tr><tr><td>e4a1b46b5c639d433d21b34b788df8d81b518729</td><td>JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 +<br/>Side Information for Face Completion: a Robust +<br/>PCA Approach +</td></tr><tr><td>e4c81c56966a763e021938be392718686ba9135e</td><td></td></tr><tr><td>e4e95b8bca585a15f13ef1ab4f48a884cd6ecfcc</td><td>Face Recognition with Independent Component Based +<br/>Super-resolution +<br/>aFaculty of Engineering and Natural Sciences, Sabanci Univ., Istanbul, Turkiye, 34956 +<br/>bSchool of Elec. and Comp. Eng. , Georgia Inst. of Tech., Atlanta, GA, USA, 30332-0250 +</td></tr><tr><td>e43ea078749d1f9b8254e0c3df4c51ba2f4eebd5</td><td>Facial Expression Recognition Based on Constrained +<br/>Local Models and Support Vector Machines +</td></tr><tr><td>e476cbcb7c1de73a7bcaeab5d0d59b8b3c4c1cbf</td><td></td></tr><tr><td>e475e857b2f5574eb626e7e01be47b416deff268</td><td>Facial Emotion Recognition Using Nonparametric +<br/>Weighted Feature Extraction and Fuzzy Classifier +</td></tr><tr><td>e4391993f5270bdbc621b8d01702f626fba36fc2</td><td>Author manuscript, published in "18th Scandinavian Conference on Image Analysis (2013)" +<br/> DOI : 10.1007/978-3-642-38886-6_31 +</td></tr><tr><td>e4d8ba577cabcb67b4e9e1260573aea708574886</td><td>UM SISTEMA DE RECOMENDAC¸ ˜AO INTELIGENTE BASEADO EM V´IDIO +<br/>AULAS PARA EDUCAC¸ ˜AO A DIST ˆANCIA +<br/>Gaspare Giuliano Elias Bruno +<br/>Tese de Doutorado apresentada ao Programa +<br/>de P´os-gradua¸c˜ao em Engenharia de Sistemas e +<br/>Computa¸c˜ao, COPPE, da Universidade Federal +<br/>do Rio de Janeiro, como parte dos requisitos +<br/>necess´arios `a obten¸c˜ao do t´ıtulo de Doutor em +<br/>Engenharia de Sistemas e Computa¸c˜ao. +<br/>Orientadores: Edmundo Albuquerque de +<br/>Souza e Silva +<br/>Rosa Maria Meri Le˜ao +<br/>Rio de Janeiro +<br/>Janeiro de 2016 +</td></tr><tr><td>e475deadd1e284428b5e6efd8fe0e6a5b83b9dcd</td><td>Accepted in Pattern Recognition Letters +<br/>Pattern Recognition Letters +<br/>journal homepage: www.elsevier.com +<br/>Are you eligible? Predicting adulthood from face images via class specific mean +<br/>autoencoder +<br/>IIIT-Delhi, New Delhi, 110020, India +<br/>Article history: +<br/>Received 15 March 2017 +</td></tr><tr><td>e4d0e87d0bd6ead4ccd39fc5b6c62287560bac5b</td><td>Implicit Video Multi-Emotion Tagging by Exploiting Multi-Expression +<br/>Relations +</td></tr><tr><td>fe9c460d5ca625402aa4d6dd308d15a40e1010fa</td><td>Neural Architecture for Temporal Emotion +<br/>Classification +<br/>Universit¨at Ulm, Neuroinformatik, Germany +</td></tr><tr><td>fe7c0bafbd9a28087e0169259816fca46db1a837</td><td></td></tr><tr><td>fe48f0e43dbdeeaf4a03b3837e27f6705783e576</td><td></td></tr><tr><td>fea83550a21f4b41057b031ac338170bacda8805</td><td>Learning a Metric Embedding +<br/>for Face Recognition +<br/>using the Multibatch Method +<br/>Orcam Ltd., Jerusalem, Israel +</td></tr><tr><td>feeb0fd0e254f38b38fe5c1022e84aa43d63f7cc</td><td>EURECOM +<br/>Multimedia Communications Department +<br/>and +<br/>Mobile Communications Department +<br/>2229, route des Crˆetes +<br/>B.P. 193 +<br/>06904 Sophia-Antipolis +<br/>FRANCE +<br/>Research Report RR-11-255 +<br/>Search Pruning with Soft Biometric Systems: +<br/>Efficiency-Reliability Tradeoff +<br/>June 1st, 2011 +<br/>Last update June 1st, 2011 +<br/>1EURECOM’s research is partially supported by its industrial members: BMW Group, Cisco, +<br/>Monaco Telecom, Orange, SAP, SFR, Sharp, STEricsson, Swisscom, Symantec, Thales. +</td></tr><tr><td>fe108803ee97badfa2a4abb80f27fa86afd9aad9</td><td></td></tr><tr><td>fe0c51fd41cb2d5afa1bc1900bbbadb38a0de139</td><td>Rahman et al. EURASIP Journal on Image and Video Processing (2015) 2015:35 +<br/>DOI 10.1186/s13640-015-0090-5 +<br/>RESEARCH +<br/>Open Access +<br/>Bayesian face recognition using 2D +<br/>Gaussian-Hermite moments +</td></tr><tr><td>c8db8764f9d8f5d44e739bbcb663fbfc0a40fb3d</td><td>Modeling for part-based visual object +<br/>detection based on local features +<br/>Von der Fakult¨at f¨ur Elektrotechnik und Informationstechnik +<br/>der Rheinisch-Westf¨alischen Technischen Hochschule Aachen +<br/>zur Erlangung des akademischen Grades eines Doktors +<br/>der Ingenieurwissenschaften genehmigte Dissertation +<br/>vorgelegt von +<br/>Diplom-Ingenieur +<br/>aus Neuss +<br/>Berichter: +<br/>Univ.-Prof. Dr.-Ing. Jens-Rainer Ohm +<br/>Univ.-Prof. Dr.-Ing. Til Aach +<br/>Tag der m¨undlichen Pr¨ufung: 28. September 2011 +<br/>Diese Dissertation ist auf den Internetseiten der +<br/>Hochschulbibliothek online verf¨ugbar. +</td></tr><tr><td>c86e6ed734d3aa967deae00df003557b6e937d3d</td><td>Generative Adversarial Networks with +<br/>Decoder-Encoder Output Noise +<br/>conditional distribution of their neighbors. In [32], Portilla and +<br/>Simoncelli proposed a parametric texture model based on joint +<br/>statistics, which uses a decomposition method that is called +<br/>steerable pyramid decomposition to decompose the texture +<br/>of images. An example-based super-resolution algorithm [11] +<br/>was proposed in 2002, which uses a Markov network to model +<br/>the spatial relationship between the pixels of an image. A +<br/>scene completion algorithm [16] was proposed in 2007, which +<br/>applied a semantic scene match technique. These traditional +<br/>algorithms can be applied to particular image generation tasks, +<br/>such as texture synthesis and super-resolution. Their common +<br/>characteristic is that they predict the images pixel by pixel +<br/>rather than generate an image as a whole, and the basic idea +<br/>of them is to make an interpolation according to the existing +<br/>part of the images. Here, the problem is, given a set of images, +<br/>can we generate totally new images with the same distribution +<br/>of the given ones? +</td></tr><tr><td>c8a4b4fe5ff2ace9ab9171a9a24064b5a91207a3</td><td>LOCATING FACIAL LANDMARKS WITH BINARY MAP CROSS-CORRELATIONS +<br/>J´er´emie Nicolle +<br/>K´evin Bailly +<br/>Univ. Pierre & Marie Curie, ISIR - CNRS UMR 7222, F-75005, Paris - France +</td></tr><tr><td>c866a2afc871910e3282fd9498dce4ab20f6a332</td><td>Noname manuscript No. +<br/>(will be inserted by the editor) +<br/>Surveillance Face Recognition Challenge +<br/>Received: date / Accepted: date +</td></tr><tr><td>c82c147c4f13e79ad49ef7456473d86881428b89</td><td></td></tr><tr><td>c84233f854bbed17c22ba0df6048cbb1dd4d3248</td><td>Exploring Locally Rigid Discriminative +<br/>Patches for Learning Relative Attributes +<br/>http://researchweb.iiit.ac.in/~yashaswi.verma/ +<br/>http://www.iiit.ac.in/~jawahar/ +<br/>CVIT +<br/>IIIT-Hyderabad, India +<br/>http://cvit.iiit.ac.in +</td></tr><tr><td>c8adbe00b5661ab9b3726d01c6842c0d72c8d997</td><td>Deep Architectures for Face Attributes +<br/>Computer Vision and Machine Learning Group, Flickr, Yahoo, +</td></tr><tr><td>fb4545782d9df65d484009558e1824538030bbb1</td><td></td></tr><tr><td>fb5280b80edcf088f9dd1da769463d48e7b08390</td><td></td></tr><tr><td>fba464cb8e3eff455fe80e8fb6d3547768efba2f</td><td> +<br/>International Journal of Engineering and Applied Sciences (IJEAS) +<br/> ISSN: 2394-3661, Volume-3, Issue-2, February 2016 +<br/>Survey Paper on Emotion Recognition +<br/> +</td></tr><tr><td>fbb2f81fc00ee0f257d4aa79bbef8cad5000ac59</td><td>Reading Hidden Emotions: Spontaneous +<br/>Micro-expression Spotting and Recognition +</td></tr><tr><td>fb9ad920809669c1b1455cc26dbd900d8e719e61</td><td>3D Gaze Estimation from Remote RGB-D Sensors +<br/>THÈSE NO 6680 (2015) +<br/>PRÉSENTÉE LE 9 OCTOBRE 2015 +<br/>À LA FACULTÉ DES SCIENCES ET TECHNIQUES DE L'INGÉNIEUR +<br/>LABORATOIRE DE L'IDIAP +<br/>PROGRAMME DOCTORAL EN GÉNIE ÉLECTRIQUE +<br/>ÉCOLE POLYTECHNIQUE FÉDÉRALE DE LAUSANNE +<br/>POUR L'OBTENTION DU GRADE DE DOCTEUR ÈS SCIENCES +<br/>PAR +<br/>acceptée sur proposition du jury: +<br/>Prof. K. Aminian, président du jury +<br/>Dr J.-M. Odobez, directeur de thèse +<br/>Prof. L.-Ph. Morency, rapporteur +<br/>Prof. D. Witzner Hansen, rapporteur +<br/>Dr R. Boulic, rapporteur +<br/>Suisse +<br/>2015 +</td></tr><tr><td>edef98d2b021464576d8d28690d29f5431fd5828</td><td>Pixel-Level Alignment of Facial Images +<br/>for High Accuracy Recognition +<br/>Using Ensemble of Patches +</td></tr><tr><td>ed04e161c953d345bcf5b910991d7566f7c486f7</td><td>Combining facial expression analysis and synthesis on a +<br/>Mirror my emotions! +<br/>robot +</td></tr><tr><td>c178a86f4c120eca3850a4915134fff44cbccb48</td><td></td></tr><tr><td>c1d2d12ade031d57f8d6a0333cbe8a772d752e01</td><td>Journal of Math-for-Industry, Vol.2(2010B-5), pp.147–156 +<br/>Convex optimization techniques for the efficient recovery of a sparsely +<br/>corrupted low-rank matrix +<br/>D 案 +<br/>Received on August 10, 2010 / Revised on August 31, 2010 +<br/>E 案 +</td></tr><tr><td>c10a15e52c85654db9c9343ae1dd892a2ac4a279</td><td>Int J Comput Vis (2012) 100:134–153 +<br/>DOI 10.1007/s11263-011-0494-3 +<br/>Learning the Relative Importance of Objects from Tagged Images +<br/>for Retrieval and Cross-Modal Search +<br/>Received: 16 December 2010 / Accepted: 23 August 2011 / Published online: 18 October 2011 +<br/>© Springer Science+Business Media, LLC 2011 +</td></tr><tr><td>c1fc70e0952f6a7587b84bf3366d2e57fc572fd7</td><td></td></tr><tr><td>c1dfabe36a4db26bf378417985a6aacb0f769735</td><td>Journal of Computer Vision and Image Processing, NWPJ-201109-50 +<br/>1 +<br/>Describing Visual Scene through EigenMaps +<br/> +</td></tr><tr><td>c1482491f553726a8349337351692627a04d5dbe</td><td></td></tr><tr><td>c1ff88493721af1940df0d00bcfeefaa14f1711f</td><td>CVPR +<br/>#1369 +<br/>000 +<br/>001 +<br/>002 +<br/>003 +<br/>004 +<br/>005 +<br/>006 +<br/>007 +<br/>008 +<br/>009 +<br/>010 +<br/>011 +<br/>012 +<br/>013 +<br/>014 +<br/>015 +<br/>016 +<br/>017 +<br/>018 +<br/>019 +<br/>020 +<br/>021 +<br/>022 +<br/>023 +<br/>024 +<br/>025 +<br/>026 +<br/>027 +<br/>028 +<br/>029 +<br/>030 +<br/>031 +<br/>032 +<br/>033 +<br/>034 +<br/>035 +<br/>036 +<br/>037 +<br/>038 +<br/>039 +<br/>040 +<br/>041 +<br/>042 +<br/>043 +<br/>044 +<br/>045 +<br/>046 +<br/>047 +<br/>048 +<br/>049 +<br/>050 +<br/>051 +<br/>052 +<br/>053 +<br/>CVPR 2010 Submission #1369. CONFIDENTIAL REVIEW COPY. DO NOT DISTRIBUTE. +<br/>CVPR +<br/>#1369 +<br/>Subspace Regression: Predicting a Subspace from one Sample +<br/>Anonymous CVPR submission +<br/>Paper ID 1369 +</td></tr><tr><td>c11eb653746afa8148dc9153780a4584ea529d28</td><td>Global and Local Consistent Wavelet-domain Age +<br/>Synthesis +</td></tr><tr><td>c1ebbdb47cb6a0ed49c4d1cf39d7565060e6a7ee</td><td>Robust Facial Landmark Localization Based on +</td></tr><tr><td>c17a332e59f03b77921942d487b4b102b1ee73b6</td><td>Learning an appearance-based gaze estimator +<br/>from one million synthesised images +<br/>Tadas Baltruˇsaitis2 +</td></tr><tr><td>c1e76c6b643b287f621135ee0c27a9c481a99054</td><td></td></tr><tr><td>c6f3399edb73cfba1248aec964630c8d54a9c534</td><td>A Comparison of CNN-based Face and Head Detectors for +<br/>Real-Time Video Surveillance Applications +<br/>1 ´Ecole de technologie sup´erieure, Universit´e du Qu´ebec, Montreal, Canada +<br/>2 Genetec Inc., Montreal, Canada +</td></tr><tr><td>c62c07de196e95eaaf614fb150a4fa4ce49588b4</td><td>Proceedings of the Twenty-Seventh International Joint Conference on Artificial Intelligence (IJCAI-18) +<br/>1078 +</td></tr><tr><td>ec1e03ec72186224b93b2611ff873656ed4d2f74</td><td>JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 +<br/>3D Reconstruction of “In-the-Wild” Faces in +<br/>Images and Videos +</td></tr><tr><td>ec22eaa00f41a7f8e45ed833812d1ac44ee1174e</td><td></td></tr><tr><td>ec54000c6c0e660dd99051bdbd7aed2988e27ab8</td><td>TWO IN ONE: JOINT POSE ESTIMATION AND FACE RECOGNITION WITH P2CA1 +<br/>*Dept. Teoria del Senyal i Comunicacions - Universitat Politècnica de Catalunya, Barcelona, Spain +<br/>+Dipartimento di Elettronica e Informazione - Politecnico di Milano, Meiland, Italy +</td></tr><tr><td>ec0104286c96707f57df26b4f0a4f49b774c486b</td><td>758 +<br/>An Ensemble CNN2ELM for Age Estimation +</td></tr><tr><td>4e32fbb58154e878dd2fd4b06398f85636fd0cf4</td><td>A Hierarchical Matcher using Local Classifier Chains +<br/>L. Zhang and I.A. Kakadiaris +<br/>Computational Biomedicine Lab, 4849 Calhoun Rd, Rm 373, Houston, TX 77204 +</td></tr><tr><td>4e27fec1703408d524d6b7ed805cdb6cba6ca132</td><td>SSD-Sface: Single shot multibox detector for small faces +<br/>C. Thuis +</td></tr><tr><td>4e6c9be0b646d60390fe3f72ce5aeb0136222a10</td><td>Long-term Temporal Convolutions +<br/>for Action Recognition +</td></tr><tr><td>4e444db884b5272f3a41e4b68dc0d453d4ec1f4c</td><td></td></tr><tr><td>4ef0a6817a7736c5641dc52cbc62737e2e063420</td><td>International Journal of Advanced Computer Research (ISSN (Print): 2249-7277 ISSN (Online): 2277-7970) +<br/>Volume-4 Number-4 Issue-17 December-2014 +<br/>Study of Face Recognition Techniques +<br/>Received: 10-November-2014; Revised: 18-December-2014; Accepted: 23-December-2014 +<br/>©2014 ACCENTS +</td></tr><tr><td>4e7ebf3c4c0c4ecc48348a769dd6ae1ebac3bf1b</td><td></td></tr><tr><td>4e0e49c280acbff8ae394b2443fcff1afb9bdce6</td><td>Automatic learning of gait signatures for people identification +<br/>F.M. Castro +<br/>Univ. of Malaga +<br/>fcastro<at>uma.es +<br/>M.J. Mar´ın-Jim´enez +<br/>Univ. of Cordoba +<br/>mjmarin<at>uco.es +<br/>N. Guil +<br/>Univ. of Malaga +<br/>nguil<at>uma.es +<br/>N. P´erez de la Blanca +<br/>Univ. of Granada +<br/>nicolas<at>ugr.es +</td></tr><tr><td>4e4e8fc9bbee816e5c751d13f0d9218380d74b8f</td><td></td></tr><tr><td>20a88cc454a03d62c3368aa1f5bdffa73523827b</td><td></td></tr><tr><td>20a432a065a06f088d96965f43d0055675f0a6c1</td><td>In: Proc. of the 25th Int. Conference on Artificial Neural Networks (ICANN) +<br/>Part II, LNCS 9887, pp. 80-87, Barcelona, Spain, September 2016 +<br/>The final publication is available at Springer via +<br/>http://dx.doi.org//10.1007/978-3-319-44781-0_10 +<br/>The Effects of Regularization on Learning Facial +<br/>Expressions with Convolutional Neural Networks +<br/><b></b><br/>Vogt-Koelln-Strasse 30, 22527 Hamburg, Germany +<br/>http://www.informatik.uni-hamburg.de/WTM +</td></tr><tr><td>20e504782951e0c2979d9aec88c76334f7505393</td><td>Robust LSTM-Autoencoders for Face De-Occlusion +<br/>in the Wild +</td></tr><tr><td>20ade100a320cc761c23971d2734388bfe79f7c5</td><td>IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE +<br/>Subspace Clustering via Good Neighbors +</td></tr><tr><td>20767ca3b932cbc7b8112db21980d7b9b3ea43a3</td><td></td></tr><tr><td>20c2a5166206e7ffbb11a23387b9c5edf42b5230</td><td></td></tr><tr><td>2098983dd521e78746b3b3fa35a22eb2fa630299</td><td></td></tr><tr><td>206e24f7d4b3943b35b069ae2d028143fcbd0704</td><td>Learning Structure and Strength of CNN Filters for Small Sample Size Training +<br/>IIIT-Delhi, India +</td></tr><tr><td>2059d2fecfa61ddc648be61c0cbc9bc1ad8a9f5b</td><td>TRANSACTIONS ON AUDIO, SPEECH, AND LANGUAGE PROCESSING, VOL. 23, NO. 4, APRIL 2015 +<br/>Co-Localization of Audio Sources in Images Using +<br/>Binaural Features and Locally-Linear Regression +<br/>∗ INRIA Grenoble Rhˆone-Alpes, Montbonnot Saint-Martin, France +<br/>† Univ. Grenoble Alpes, GIPSA-Lab, France +<br/>‡ Dept. Electrical Eng., Technion-Israel Inst. of Technology, Haifa, Israel +</td></tr><tr><td>206fbe6ab6a83175a0ef6b44837743f8d5f9b7e8</td><td></td></tr><tr><td>20111924fbf616a13d37823cd8712a9c6b458cd6</td><td>International Journal of Computer Applications (0975 – 8887) +<br/>Volume 130 – No.11, November2015 +<br/>Linear Regression Line based Partial Face Recognition +<br/>Naveena M. +<br/>Department of Studies in +<br/>Computer Science, +<br/>Manasagagothri, +<br/>Mysore. +<br/>Department of Studies in +<br/>Computer Science, +<br/>Manasagagothri, +<br/>Mysore. +<br/>P. Nagabhushan +<br/>Department of Studies in +<br/>Computer Science, +<br/>Manasagagothri, +<br/>Mysore. +<br/>images. In +</td></tr><tr><td>20532b1f80b509f2332b6cfc0126c0f80f438f10</td><td>A deep matrix factorization method for learning +<br/>attribute representations +<br/>Bj¨orn W. Schuller, Senior member, IEEE +</td></tr><tr><td>205af28b4fcd6b569d0241bb6b255edb325965a4</td><td>Intel Serv Robotics (2008) 1:143–157 +<br/>DOI 10.1007/s11370-007-0014-z +<br/>SPECIAL ISSUE +<br/>Facial expression recognition and tracking for intelligent human-robot +<br/>interaction +<br/>Received: 27 June 2007 / Accepted: 6 December 2007 / Published online: 23 January 2008 +<br/>© Springer-Verlag 2008 +</td></tr><tr><td>20a0b23741824a17c577376fdd0cf40101af5880</td><td>Learning to track for spatio-temporal action localization +<br/>Zaid Harchaouia,b +<br/>b NYU +<br/>a Inria∗ +</td></tr><tr><td>18c72175ddbb7d5956d180b65a96005c100f6014</td><td>IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 23, NO. 6, +<br/>JUNE 2001 +<br/>643 +<br/>From Few to Many: Illumination Cone +<br/>Models for Face Recognition under +<br/>Variable Lighting and Pose +</td></tr><tr><td>18636347b8741d321980e8f91a44ee054b051574</td><td>978-1-4244-5654-3/09/$26.00 ©2009 IEEE +<br/>37 +<br/>ICIP 2009 +</td></tr><tr><td>18206e1b988389eaab86ef8c852662accf3c3663</td><td></td></tr><tr><td>181045164df86c72923906aed93d7f2f987bce6c</td><td>RHEINISCH-WESTFÄLISCHE TECHNISCHE +<br/>HOCHSCHULE AACHEN +<br/>KNOWLEDGE-BASED SYSTEMS GROUP +<br/>Detection and Recognition of Human +<br/>Faces using Random Forests for a +<br/>Mobile Robot +<br/>MASTER OF SCIENCE THESIS +<br/>MATRICULATION NUMBER: 26 86 51 +<br/>SUPERVISOR: +<br/>SECOND SUPERVISOR: +<br/>PROF. ENRICO BLANZIERI, PH. D. +<br/>ADVISERS: +</td></tr><tr><td>18d5b0d421332c9321920b07e0e8ac4a240e5f1f</td><td>Collaborative Representation Classification +<br/>Ensemble for Face Recognition +</td></tr><tr><td>18d51a366ce2b2068e061721f43cb798177b4bb7</td><td>Cognition and Emotion +<br/>ISSN: 0269-9931 (Print) 1464-0600 (Online) Journal homepage: http://www.tandfonline.com/loi/pcem20 +<br/>Looking into your eyes: observed pupil size +<br/>influences approach-avoidance responses +<br/>eyes: observed pupil size influences approach-avoidance responses, Cognition and Emotion, DOI: +<br/>10.1080/02699931.2018.1472554 +<br/>To link to this article: https://doi.org/10.1080/02699931.2018.1472554 +<br/>View supplementary material +<br/>Published online: 11 May 2018. +<br/>Submit your article to this journal +<br/>View related articles +<br/>View Crossmark data +<br/>Full Terms & Conditions of access and use can be found at +<br/>http://www.tandfonline.com/action/journalInformation?journalCode=pcem20 +</td></tr><tr><td>1885acea0d24e7b953485f78ec57b2f04e946eaf</td><td>Combining Local and Global Features for 3D Face Tracking +<br/>Megvii (face++) Research +</td></tr><tr><td>184750382fe9b722e78d22a543e852a6290b3f70</td><td></td></tr><tr><td>18a849b1f336e3c3b7c0ee311c9ccde582d7214f</td><td>Int J Comput Vis +<br/>DOI 10.1007/s11263-012-0564-1 +<br/>Efficiently Scaling up Crowdsourced Video Annotation +<br/>A Set of Best Practices for High Quality, Economical Video Labeling +<br/>Received: 31 October 2011 / Accepted: 20 August 2012 +<br/>© Springer Science+Business Media, LLC 2012 +</td></tr><tr><td>1886b6d9c303135c5fbdc33e5f401e7fc4da6da4</td><td>Knowledge Guided Disambiguation for Large-Scale +<br/>Scene Classification with Multi-Resolution CNNs +</td></tr><tr><td>1888bf50fd140767352158c0ad5748b501563833</td><td>PA R T 1 +<br/>THE BASICS +</td></tr><tr><td>185360fe1d024a3313042805ee201a75eac50131</td><td>299 +<br/>Person De-Identification in Videos +</td></tr><tr><td>18dfc2434a95f149a6cbb583cca69a98c9de9887</td><td></td></tr><tr><td>27d709f7b67204e1e5e05fe2cfac629afa21699d</td><td></td></tr><tr><td>275b5091c50509cc8861e792e084ce07aa906549</td><td>Institut für Informatik +<br/>der Technischen +<br/>Universität München +<br/>Dissertation +<br/>Leveraging the User’s Face as a Known Object +<br/>in Handheld Augmented Reality +<br/>Sebastian Bernhard Knorr +</td></tr><tr><td>270733d986a1eb72efda847b4b55bc6ba9686df4</td><td>We are IntechOpen, +<br/>the first native scientific +<br/>publisher of Open Access books +<br/>3,350 +<br/>108,000 +<br/>1.7 M +<br/>Open access books available +<br/>International authors and editors +<br/>Downloads +<br/>Our authors are among the +<br/>151 +<br/>Countries delivered to +<br/>TOP 1% +<br/>12.2% +<br/>most cited scientists +<br/>Contributors from top 500 universities +<br/>Selection of our books indexed in the Book Citation Index +<br/>in Web of Science™ Core Collection (BKCI) +<br/>Interested in publishing with us? +<br/>Numbers displayed above are based on latest data collected. +<br/>For more information visit www.intechopen.com +</td></tr><tr><td>27da432cf2b9129dce256e5bf7f2f18953eef5a5</td><td></td></tr><tr><td>2770b095613d4395045942dc60e6c560e882f887</td><td>GridFace: Face Rectification via Learning Local +<br/>Homography Transformations +<br/>Face++, Megvii Inc. +</td></tr><tr><td>27cccf992f54966feb2ab4831fab628334c742d8</td><td>International Journal of Computer Applications (0975 – 8887) +<br/>Volume 64– No.18, February 2013 +<br/>Facial Expression Recognition by Statistical, Spatial +<br/>Features and using Decision Tree +<br/>Assistant Professor +<br/>CSIT Department +<br/>GGV BIlaspur, Chhattisgarh +<br/>India +<br/>Assistant Professor +<br/>Electronics (ECE) Department +<br/>JECRC Jaipur, Rajasthan India +<br/>IshanBhardwaj +<br/>Student of Ph.D. +<br/>Electrical Department +<br/>NIT Raipur, Chhattisgarh India +</td></tr><tr><td>27f8b01e628f20ebfcb58d14ea40573d351bbaad</td><td>DEPARTMENT OF INFORMATION ENGINEERING AND COMPUTER SCIENCE +<br/>ICT International Doctoral School +<br/>Events based Multimedia Indexing +<br/>and Retrieval +<br/>SUBMITTED TO THE DEPARTMENT OF +<br/>INFORMATION ENGINEERING AND COMPUTER SCIENCE (DISI) +<br/>IN THE PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE +<br/>OF +<br/>DOCTOR OF PHILOSOPHY +<br/>Advisor: +<br/>Examiners: Prof. Marco Carli, Universit`a degli Studi di Roma Tre, Italy +<br/>Prof. Nicola Conci, Universit`a degli Studi di Trento, Italy +<br/>Prof. Pietro Zanuttigh, Universit`a degli Studi di Padova, Italy +<br/>Prof. Giulia Boato, Universit`a degli Studi di Trento, Italy +<br/>December 2017 +</td></tr><tr><td>274f87ad659cd90382ef38f7c6fafc4fc7f0d74d</td><td></td></tr><tr><td>27ee8482c376ef282d5eb2e673ab042f5ded99d7</td><td>Scale Normalization for the Distance Maps AAM. +<br/>Avenue de la boulaie, BP 81127, +<br/>35 511 Cesson-S´evign´e, France +<br/>Sup´elec, IETR-SCEE Team +</td></tr><tr><td>4b89cf7197922ee9418ae93896586c990e0d2867</td><td>LATEX Author Guidelines for CVPR Proceedings +<br/>First Author +<br/>Institution1 +<br/>Institution1 address +</td></tr><tr><td>4b04247c7f22410681b6aab053d9655cf7f3f888</td><td>Robust Face Recognition by Constrained Part-based +<br/>Alignment +</td></tr><tr><td>4b60e45b6803e2e155f25a2270a28be9f8bec130</td><td>Attribute Based Object Identification +</td></tr><tr><td>4b48e912a17c79ac95d6a60afed8238c9ab9e553</td><td>JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 +<br/>Minimum Margin Loss for Deep Face Recognition +</td></tr><tr><td>4b5eeea5dd8bd69331bd4bd4c66098b125888dea</td><td>Human Activity Recognition Using Conditional +<br/>Random Fields and Privileged Information +<br/>submitted to +<br/>the designated by the General Assembly Composition of the +<br/>Department of Computer Science & Engineering Inquiry +<br/>Committee +<br/>by +<br/>in partial fulfillment of the Requirements for the Degree of +<br/>DOCTOR OF PHILOSOPHY +<br/>February 2016 +</td></tr><tr><td>4bbbee93519a4254736167b31be69ee1e537f942</td><td></td></tr><tr><td>4b6be933057d939ddfa665501568ec4704fabb39</td><td></td></tr><tr><td>4be03fd3a76b07125cd39777a6875ee59d9889bd</td><td>CONTENT-BASED ANALYSIS FOR ACCESSING AUDIOVISUAL ARCHIVES: +<br/>ALTERNATIVES FOR CONCEPT-BASED INDEXING AND SEARCH +<br/>ESAT/PSI - IBBT +<br/>KU Leuven, Belgium +</td></tr><tr><td>113e5678ed8c0af2b100245057976baf82fcb907</td><td>Facing Imbalanced Data +<br/>Recommendations for the Use of Performance Metrics +</td></tr><tr><td>11f17191bf74c80ad0b16b9f404df6d03f7c8814</td><td>Recognition of Visually Perceived Compositional +<br/>Human Actions by Multiple Spatio-Temporal Scales +<br/>Recurrent Neural Networks +</td></tr><tr><td>11367581c308f4ba6a32aac1b4a7cdb32cd63137</td><td></td></tr><tr><td>1198572784788a6d2c44c149886d4e42858d49e4</td><td>Learning Discriminative Features using Encoder/Decoder type Deep +<br/>Neural Nets +</td></tr><tr><td>11fe6d45aa2b33c2ec10d9786a71c15ec4d3dca8</td><td>970 +<br/>JUNE 2008 +<br/>Tied Factor Analysis for Face Recognition +<br/>across Large Pose Differences +</td></tr><tr><td>112780a7fe259dc7aff2170d5beda50b2bfa7bda</td><td></td></tr><tr><td>111a9645ad0108ad472b2f3b243ed3d942e7ff16</td><td>Facial Expression Classification Using +<br/>Combined Neural Networks +<br/>DEE/PUC-Rio, Marquês de São Vicente 225, Rio de Janeiro – RJ - Brazil +</td></tr><tr><td>111d0b588f3abbbea85d50a28c0506f74161e091</td><td>International Journal of Computer Applications (0975 – 8887) +<br/>Volume 134 – No.10, January 2016 +<br/>Facial Expression Recognition from Visual Information +<br/>using Curvelet Transform +<br/>Surabhi Group of Institution Bhopal +<br/>systems. Further applications +</td></tr><tr><td>7de6e81d775e9cd7becbfd1bd685f4e2a5eebb22</td><td>Labeled Faces in the Wild: A Survey +</td></tr><tr><td>7d73adcee255469aadc5e926066f71c93f51a1a5</td><td>978-1-4799-9988-0/16/$31.00 ©2016 IEEE +<br/>1283 +<br/>ICASSP 2016 +</td></tr><tr><td>7dffe7498c67e9451db2d04bb8408f376ae86992</td><td>LEAR-INRIA submission for the THUMOS workshop +<br/>LEAR, INRIA, France +</td></tr><tr><td>7d3f6dd220bec883a44596ddec9b1f0ed4f6aca2</td><td>2106 +<br/>Linear Regression for Face Recognition +</td></tr><tr><td>29ce6b54a87432dc8371f3761a9568eb3c5593b0</td><td>Kent Academic Repository +<br/>Full text document (pdf) +<br/>Citation for published version +<br/>Yassin, DK H. PHM and Hoque, Sanaul and Deravi, Farzin (2013) Age Sensitivity of Face Recognition +<br/> pp. 12-15. +<br/>DOI +<br/>https://doi.org/10.1109/EST.2013.8 +<br/>Link to record in KAR +<br/>http://kar.kent.ac.uk/43222/ +<br/>Document Version +<br/>Author's Accepted Manuscript +<br/>Copyright & reuse +<br/>Content in the Kent Academic Repository is made available for research purposes. Unless otherwise stated all +<br/>content is protected by copyright and in the absence of an open licence (eg Creative Commons), permissions +<br/>for further reuse of content should be sought from the publisher, author or other copyright holder. +<br/>Versions of research +<br/>The version in the Kent Academic Repository may differ from the final published version. +<br/>Users are advised to check http://kar.kent.ac.uk for the status of the paper. Users should always cite the +<br/>published version of record. +<br/>Enquiries +<br/>For any further enquiries regarding the licence status of this document, please contact: +<br/>If you believe this document infringes copyright then please contact the KAR admin team with the take-down +<br/>information provided at http://kar.kent.ac.uk/contact.html +</td></tr><tr><td>292eba47ef77495d2613373642b8372d03f7062b</td><td>Deep Secure Encoding: An Application to Face Recognition +</td></tr><tr><td>29e96ec163cb12cd5bd33bdf3d32181c136abaf9</td><td>Report No. UIUCDCS-R-2006-2748 +<br/>UILU-ENG-2006-1788 +<br/>Regularized Locality Preserving Projections with Two-Dimensional +<br/>Discretized Laplacian Smoothing +<br/>by +<br/>July 2006 +</td></tr><tr><td>29c1f733a80c1e07acfdd228b7bcfb136c1dff98</td><td></td></tr><tr><td>29f27448e8dd843e1c4d2a78e01caeaea3f46a2d</td><td></td></tr><tr><td>294d1fa4e1315e1cf7cc50be2370d24cc6363a41</td><td>2008 SPIE Digital Library -- Subscriber Archive Copy +</td></tr><tr><td>29d414bfde0dfb1478b2bdf67617597dd2d57fc6</td><td>Multidim Syst Sign Process (2010) 21:213–229 +<br/>DOI 10.1007/s11045-009-0099-y +<br/>Perfect histogram matching PCA for face recognition +<br/>Received: 10 August 2009 / Revised: 21 November 2009 / Accepted: 29 December 2009 / +<br/>Published online: 14 January 2010 +<br/>© Springer Science+Business Media, LLC 2010 +</td></tr><tr><td>290136947fd44879d914085ee51d8a4f433765fa</td><td>On a Taxonomy of Facial Features +</td></tr><tr><td>2957715e96a18dbb5ed5c36b92050ec375214aa6</td><td>Improving Face Attribute Detection with Race and Gender Diversity +<br/>InclusiveFaceNet: +</td></tr><tr><td>291265db88023e92bb8c8e6390438e5da148e8f5</td><td>MS-Celeb-1M: A Dataset and Benchmark for +<br/>Large-Scale Face Recognition +<br/>Microsoft Research +</td></tr><tr><td>2921719b57544cfe5d0a1614d5ae81710ba804fa</td><td>Face Recognition Enhancement Based on Image +<br/>File Formats and Wavelet De-noising +<br/> +</td></tr><tr><td>29a013b2faace976f2c532533bd6ab4178ccd348</td><td>This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. +<br/>Hierarchical Manifold Learning With Applications +<br/>to Supervised Classification for High-Resolution +<br/>Remotely Sensed Images +</td></tr><tr><td>29756b6b16d7b06ea211f21cdaeacad94533e8b4</td><td>Thresholding Approach based on GPU for Facial +<br/>Expression Recognition +<br/>1 Benemérita Universidad Autónoma de Puebla, Faculty of Computer Science, Puebla, México +<br/>2Instituto Tecnológico de Puebla, Puebla, México +</td></tr><tr><td>293193d24d5c4d2975e836034bbb2329b71c4fe7</td><td>Building a Corpus of Facial Expressions +<br/>for Learning-Centered Emotions +<br/>Instituto Tecnológico de Culiacán, Culiacán, Sinaloa, +<br/>Mexico +</td></tr><tr><td>2988f24908e912259d7a34c84b0edaf7ea50e2b3</td><td>A Model of Brightness Variations Due to +<br/>Illumination Changes and Non-rigid Motion +<br/>Using Spherical Harmonics +<br/>Jos´e M. Buenaposada +<br/>Dep. Ciencias de la Computaci´on, +<br/>U. Rey Juan Carlos, Spain +<br/>http://www.dia.fi.upm.es/~pcr +<br/>Inst. for Systems and Robotics +<br/>Inst. Superior T´ecnico, Portugal +<br/>http://www.isr.ist.utl.pt/~adb +<br/>Enrique Mu˜noz +<br/>Facultad de Inform´atica, +<br/>U. Complutense de Madrid, Spain +<br/>Dep. de Inteligencia Artificial, +<br/>U. Polit´ecnica de Madrid, Spain +<br/>http://www.dia.fi.upm.es/~pcr +<br/>http://www.dia.fi.upm.es/~pcr +</td></tr><tr><td>29156e4fe317b61cdcc87b0226e6f09e416909e0</td><td></td></tr><tr><td>293ade202109c7f23637589a637bdaed06dc37c9</td><td></td></tr><tr><td>7c7ab59a82b766929defd7146fd039b89d67e984</td><td>Improving Multiview Face Detection with +<br/>Multi-Task Deep Convolutional Neural Networks +<br/>Microsoft Research +<br/>One Microsoft Way, Redmond WA 98052 +</td></tr><tr><td>7c45b5824645ba6d96beec17ca8ecfb22dfcdd7f</td><td>News image annotation on a large parallel text-image corpus +<br/>Universit´e de Rennes 1/IRISA, CNRS/IRISA, INRIA Rennes-Bretagne Atlantique +<br/>Campus de Beaulieu +<br/>35042 Rennes Cedex, France +</td></tr><tr><td>7c0a6824b556696ad7bdc6623d742687655852db</td><td>18th Telecommunications forum TELFOR 2010 +<br/>Serbia, Belgrade, November 23-25, 2010. +<br/>MPCA+DATER: A Novel Approach for Face +<br/>Recognition Based on Tensor Objects +<br/>Ali. A. Shams Baboli, Member, IEEE, G. Rezai-rad, Member, IEEE, Aref. Shams Baboli +</td></tr><tr><td>7c95449a5712aac7e8c9a66d131f83a038bb7caa</td><td>This is an author produced version of Facial first impressions from another angle: How +<br/>social judgements are influenced by changeable and invariant facial properties. +<br/>White Rose Research Online URL for this paper: +<br/>http://eprints.whiterose.ac.uk/102935/ +<br/>Article: +<br/>Rhodes (2017) Facial first impressions from another angle: How social judgements are +<br/>influenced by changeable and invariant facial properties. British journal of psychology. pp. +<br/>397-415. ISSN 0007-1269 +<br/>https://doi.org/10.1111/bjop.12206 +<br/>promoting access to +<br/>White Rose research papers +<br/>http://eprints.whiterose.ac.uk/ +</td></tr><tr><td>7c3e09e0bd992d3f4670ffacb4ec3a911141c51f</td><td>Noname manuscript No. +<br/>(will be inserted by the editor) +<br/>Transferring Object-Scene Convolutional Neural Networks for +<br/>Event Recognition in Still Images +<br/>Received: date / Accepted: date +</td></tr><tr><td>7c7b0550ec41e97fcfc635feffe2e53624471c59</td><td>1051-4651/14 $31.00 © 2014 IEEE +<br/>DOI 10.1109/ICPR.2014.124 +<br/>660 +</td></tr><tr><td>7ce03597b703a3b6754d1adac5fbc98536994e8f</td><td></td></tr><tr><td>7c9a65f18f7feb473e993077d087d4806578214e</td><td>SpringerLink - Zeitschriftenbeitrag +<br/>http://www.springerlink.com/content/93hr862660nl1164/?p=abe5352... +<br/>Deutsch +<br/>Deutsch +<br/>Go +<br/>Vorherige Beitrag Nächste Beitrag +<br/>Beitrag markieren +<br/>In den Warenkorb legen +<br/>Zu gespeicherten Artikeln +<br/>hinzufügen +<br/>Permissions & Reprints +<br/>Diesen Artikel empfehlen +<br/>Ergebnisse +<br/>finden +<br/>Erweiterte Suche +<br/>Go +<br/>im gesamten Inhalt +<br/>in dieser Zeitschrift +<br/>in diesem Heft +<br/>Diesen Beitrag exportieren +<br/>Diesen Beitrag exportieren als RIS +<br/>| Text +<br/>Text +<br/>PDF +<br/>PDF ist das gebräuchliche Format +<br/>für Online Publikationen. Die Größe +<br/>dieses Dokumentes beträgt 564 +<br/>Kilobyte. Je nach Art Ihrer +<br/>Internetverbindung kann der +<br/>Download einige Zeit in Anspruch +<br/>nehmen. +<br/>öffnen: Gesamtdokument +<br/>Publikationsart Subject Collections +<br/>Zurück zu: Journal Issue +<br/>Athens Authentication Point +<br/>Zeitschriftenbeitrag +<br/>Willkommen! +<br/>Um unsere personalisierten +<br/>Angebote nutzen zu können, +<br/>müssen Sie angemeldet sein. +<br/>Login +<br/>Jetzt registrieren +<br/>Zugangsdaten vergessen? +<br/>Hilfe. +<br/>Mein Menü +<br/>Markierte Beiträge +<br/>Alerts +<br/>Meine Bestellungen +<br/>Private emotions versus social interaction: a data-driven approach towards +<br/>analysing emotion in speech +<br/>Zeitschrift +<br/>Verlag +<br/>ISSN +<br/>Heft +<br/>Kategorie +<br/>DOI +<br/>Seiten +<br/>Subject Collection +<br/>SpringerLink Date +<br/>User Modeling and User-Adapted Interaction +<br/>Springer Netherlands +<br/>0924-1868 (Print) 1573-1391 (Online) +<br/>Volume 18, Numbers 1-2 / Februar 2008 +<br/>Original Paper +<br/>10.1007/s11257-007-9039-4 +<br/>175-206 +<br/>Informatik +<br/>Freitag, 12. Oktober 2007 +<br/>Gespeicherte Beiträge +<br/>Alle +<br/>Favoriten +<br/>(1) Lehrstuhl für Mustererkennung, FAU Erlangen – Nürnberg, Martensstr. 3, 91058 Erlangen, +<br/>Germany +<br/>Received: 3 July 2006 Accepted: 14 January 2007 Published online: 12 October 2007 +</td></tr><tr><td>7c1e1c767f7911a390d49bed4f73952df8445936</td><td>NON-RIGID OBJECT DETECTION WITH LOCAL INTERLEAVED SEQUENTIAL ALIGNMENT (LISA) +<br/>Non-Rigid Object Detection with Local +<br/>Interleaved Sequential Alignment (LISA) +<br/>and Tom´aˇs Svoboda, Member, IEEE +</td></tr><tr><td>7c349932a3d083466da58ab1674129600b12b81c</td><td></td></tr><tr><td>1648cf24c042122af2f429641ba9599a2187d605</td><td>Boosting Cross-Age Face Verification via Generative Age Normalization +<br/>(cid:2) Orange Labs, 4 rue Clos Courtel, 35512 Cesson-S´evign´e, France +<br/>† Eurecom, 450 route des Chappes, 06410 Biot, France +</td></tr><tr><td>162403e189d1b8463952fa4f18a291241275c354</td><td>Action Recognition with Spatio-Temporal +<br/>Visual Attention on Skeleton Image Sequences +<br/>With a strong ability of modeling sequential data, Recur- +<br/>rent Neural Networks (RNN) with Long Short-Term Memory +<br/>(LSTM) neurons outperform the previous hand-crafted feature +<br/>based methods [9], [10]. Each skeleton frame is converted into +<br/>a feature vector and the whole sequence is fed into the RNN. +<br/>Despite the strong ability in modeling temporal sequences, +<br/>RNN structures lack the ability to efficiently learn the spatial +<br/>relations between the joints. To better use spatial information, +<br/>a hierarchical structure is proposed in [11], [12] that feeds +<br/>the joints into the network as several pre-defined body part +<br/>groups. However, +<br/>limit +<br/>the effectiveness of representing spatial relations. A spatio- +<br/>temporal 2D LSTM (ST-LSTM) network [13] is proposed +<br/>to learn the spatial and temporal relations simultaneously. +<br/>Furthermore, a two-stream RNN structure [14] is proposed to +<br/>learn the spatio-temporal relations with two RNN branches. +<br/>the pre-defined body regions still +</td></tr><tr><td>160259f98a6ec4ec3e3557de5e6ac5fa7f2e7f2b</td><td>Discriminant Multi-Label Manifold Embedding for Facial Action Unit +<br/>Detection +<br/>Signal Procesing Laboratory (LTS5), ´Ecole Polytechnique F´ed´erale de Lausanne, Switzerland +</td></tr><tr><td>16671b2dc89367ce4ed2a9c241246a0cec9ec10e</td><td>2006 +<br/>Detecting the Number of Clusters +<br/>in n-Way Probabilistic Clustering +</td></tr><tr><td>16de1324459fe8fdcdca80bba04c3c30bb789bdf</td><td></td></tr><tr><td>16892074764386b74b6040fe8d6946b67a246a0b</td><td></td></tr><tr><td>16395b40e19cbc6d5b82543039ffff2a06363845</td><td>Action Recognition in Video Using Sparse Coding and Relative Features +<br/>Anal´ı Alfaro +<br/>P. Universidad Catolica de Chile +<br/>P. Universidad Catolica de Chile +<br/>P. Universidad Catolica de Chile +<br/>Santiago, Chile +<br/>Santiago, Chile +<br/>Santiago, Chile +</td></tr><tr><td>16286fb0f14f6a7a1acc10fcd28b3ac43f12f3eb</td><td>J Nonverbal Behav +<br/>DOI 10.1007/s10919-008-0059-5 +<br/>O R I G I N A L P A P E R +<br/>All Smiles are Not Created Equal: Morphology +<br/>and Timing of Smiles Perceived as Amused, Polite, +<br/>and Embarrassed/Nervous +<br/>Ó Springer Science+Business Media, LLC 2008 +</td></tr><tr><td>166186e551b75c9b5adcc9218f0727b73f5de899</td><td>Volume 4, Issue 2, February 2016 +<br/>International Journal of Advance Research in +<br/>Computer Science and Management Studies +<br/>Research Article / Survey Paper / Case Study +<br/>Available online at: www.ijarcsms.com +<br/>ISSN: 2321-7782 (Online) +<br/>Automatic Age and Gender Recognition in Human Face Image +<br/>Dataset using Convolutional Neural Network System +<br/>Subhani Shaik1 +<br/>Assoc. Prof & Head of the Department +<br/>Department of CSE, +<br/>Associate Professor +<br/>Department of CSE, +<br/>St.Mary’s Group of Institutions Guntur +<br/>St.Mary’s Group of Institutions Guntur +<br/>Chebrolu(V&M),Guntur(Dt), +<br/>Andhra Pradesh - India +<br/>Chebrolu(V&M),Guntur(Dt), +<br/>Andhra Pradesh - India +</td></tr><tr><td>16d9b983796ffcd151bdb8e75fc7eb2e31230809</td><td>EUROGRAPHICS 2018 / D. Gutierrez and A. Sheffer +<br/>(Guest Editors) +<br/>Volume 37 (2018), Number 2 +<br/>GazeDirector: Fully Articulated Eye Gaze Redirection in Video +<br/>ID: paper1004 +</td></tr><tr><td>1679943d22d60639b4670eba86665371295f52c3</td><td></td></tr><tr><td>169076ffe5e7a2310e98087ef7da25aceb12b62d</td><td></td></tr><tr><td>161eb88031f382e6a1d630cd9a1b9c4bc6b47652</td><td>1 +<br/>Automatic Facial Expression Recognition +<br/>Using Features of Salient Facial Patches +</td></tr><tr><td>4209783b0cab1f22341f0600eed4512155b1dee6</td><td>Accurate and Efficient Similarity Search for Large Scale Face Recognition +<br/>BUPT +<br/>BUPT +<br/>BUPT +</td></tr><tr><td>42e3dac0df30d754c7c7dab9e1bb94990034a90d</td><td>PANDA: Pose Aligned Networks for Deep Attribute Modeling +<br/>2EECS, UC Berkeley +<br/>1Facebook AI Research +</td></tr><tr><td>429c3588ce54468090cc2cf56c9b328b549a86dc</td><td></td></tr><tr><td>42cc9ea3da1277b1f19dff3d8007c6cbc0bb9830</td><td>Coordinated Local Metric Learning +<br/>Inria∗ +</td></tr><tr><td>42350e28d11e33641775bef4c7b41a2c3437e4fd</td><td>212 +<br/>Multilinear Discriminant Analysis +<br/>for Face Recognition +</td></tr><tr><td>42e155ea109eae773dadf74d713485be83fca105</td><td></td></tr><tr><td>4270460b8bc5299bd6eaf821d5685c6442ea179a</td><td>Int J Comput Vis (2009) 84: 163–183 +<br/>DOI 10.1007/s11263-008-0147-3 +<br/>Partial Similarity of Objects, or How to Compare a Centaur +<br/>to a Horse +<br/>Received: 30 September 2007 / Accepted: 3 June 2008 / Published online: 26 July 2008 +<br/>© Springer Science+Business Media, LLC 2008 +</td></tr><tr><td>429d4848d03d2243cc6a1b03695406a6de1a7abd</td><td>Face Recognition based on Logarithmic Fusion +<br/>International Journal of Soft Computing and Engineering (IJSCE) +<br/>ISSN: 2231-2307, Volume-2, Issue-3, July 2012 +<br/>of SVD and KT +<br/>Ramachandra A C, Raja K B, Venugopal K R, L M Patnaik +<br/>to +<br/> +</td></tr><tr><td>424259e9e917c037208125ccc1a02f8276afb667</td><td></td></tr><tr><td>42ecfc3221c2e1377e6ff849afb705ecd056b6ff</td><td>Pose Invariant Face Recognition under Arbitrary +<br/>Unknown Lighting using Spherical Harmonics +<br/>Department of Computer Science, +<br/>SUNY at Stony Brook, NY, 11790 +</td></tr><tr><td>421955c6d2f7a5ffafaf154a329a525e21bbd6d3</td><td>570 +<br/>IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 22, NO. 6, +<br/>JUNE 2000 +<br/>Evolutionary Pursuit and Its +<br/>Application to Face Recognition +</td></tr><tr><td>42e0127a3fd6a96048e0bc7aab6d0ae88ba00fb0</td><td></td></tr><tr><td>42df75080e14d32332b39ee5d91e83da8a914e34</td><td>4280 +<br/>Illumination Compensation Using Oriented +<br/>Local Histogram Equalization and +<br/>Its Application to Face Recognition +</td></tr><tr><td>89945b7cd614310ebae05b8deed0533a9998d212</td><td>Divide-and-Conquer Method for L1 Norm Matrix +<br/>Factorization in the Presence of Outliers and +<br/>Missing Data +</td></tr><tr><td>89de30a75d3258816c2d4d5a733d2bef894b66b9</td><td></td></tr><tr><td>8913a5b7ed91c5f6dec95349fbc6919deee4fc75</td><td>BigBIRD: A Large-Scale 3D Database of Object Instances +</td></tr><tr><td>89d3a57f663976a9ac5e9cdad01267c1fc1a7e06</td><td>Neural Class-Specific Regression for face +<br/>verification +</td></tr><tr><td>891b10c4b3b92ca30c9b93170ec9abd71f6099c4</td><td>Facial landmark detection using structured output deep +<br/>neural networks +<br/>Soufiane Belharbi ∗1, Cl´ement Chatelain∗1, Romain H´erault∗1, and S´ebastien +<br/>1LITIS EA 4108, INSA de Rouen, Saint ´Etienne du Rouvray 76800, France +<br/>2LITIS EA 4108, UFR des Sciences, Universit´e de Rouen, France. +<br/>September 24, 2015 +</td></tr><tr><td>45c340c8e79077a5340387cfff8ed7615efa20fd</td><td></td></tr><tr><td>45e7ddd5248977ba8ec61be111db912a4387d62f</td><td>CHEN ET AL.: ADVERSARIAL POSENET +<br/>Adversarial Learning of Structure-Aware Fully +<br/>Convolutional Networks for Landmark +<br/>Localization +</td></tr><tr><td>45f3bf505f1ce9cc600c867b1fb2aa5edd5feed8</td><td></td></tr><tr><td>4560491820e0ee49736aea9b81d57c3939a69e12</td><td>Investigating the Impact of Data Volume and +<br/>Domain Similarity on Transfer Learning +<br/>Applications +<br/>State Farm Insurance, Bloomington IL 61710, USA, +</td></tr><tr><td>4571626d4d71c0d11928eb99a3c8b10955a74afe</td><td>Geometry Guided Adversarial Facial Expression Synthesis +<br/>1National Laboratory of Pattern Recognition, CASIA +<br/>2Center for Research on Intelligent Perception and Computing, CASIA +<br/>3Center for Excellence in Brain Science and Intelligence Technology, CAS +</td></tr><tr><td>4534d78f8beb8aad409f7bfcd857ec7f19247715</td><td>Under review as a conference paper at ICLR 2017 +<br/>TRANSFORMATION-BASED MODELS OF VIDEO +<br/>SEQUENCES +<br/>Facebook AI Research +</td></tr><tr><td>459e840ec58ef5ffcee60f49a94424eb503e8982</td><td>One-shot Face Recognition by Promoting Underrepresented Classes +<br/>Microsoft +<br/>One Microsoft Way, Redmond, Washington, United States +</td></tr><tr><td>45fbeed124a8956477dbfc862c758a2ee2681278</td><td></td></tr><tr><td>451c42da244edcb1088e3c09d0f14c064ed9077e</td><td>1964 +<br/>© EURASIP, 2011 - ISSN 2076-1465 +<br/>19th European Signal Processing Conference (EUSIPCO 2011) +<br/>INTRODUCTION +</td></tr><tr><td>4511e09ee26044cb46073a8c2f6e1e0fbabe33e8</td><td></td></tr><tr><td>45a6333fc701d14aab19f9e2efd59fe7b0e89fec</td><td>HAND POSTURE DATASET CREATION FOR GESTURE +<br/>RECOGNITION +<br/>Luis Anton-Canalis +<br/>Instituto de Sistemas Inteligentes y Aplicaciones Numericas en Ingenieria +<br/>Campus Universitario de Tafira, 35017 Gran Canaria, Spain +<br/>Elena Sanchez-Nielsen +<br/>Departamento de E.I.O. y Computacion +<br/>38271 Universidad de La Laguna, Spain +<br/>Keywords: +<br/>Image understanding, Gesture recognition, Hand dataset. +</td></tr><tr><td>1ffe20eb32dbc4fa85ac7844178937bba97f4bf0</td><td>Face Clustering: Representation and Pairwise +<br/>Constraints +</td></tr><tr><td>1f8304f4b51033d2671147b33bb4e51b9a1e16fe</td><td>Noname manuscript No. +<br/>(will be inserted by the editor) +<br/>Beyond Trees: +<br/>MAP Inference in MRFs via Outer-Planar Decomposition +<br/>Received: date / Accepted: date +</td></tr><tr><td>1f9ae272bb4151817866511bd970bffb22981a49</td><td>An Iterative Regression Approach for Face Pose Estima- +<br/>tion from RGB Images +<br/>This paper presents a iterative optimization method, explicit shape regression, for face pose +<br/>detection and localization. The regression function is learnt to find out the entire facial shape +<br/>and minimize the alignment errors. A cascaded learning framework is employed to enhance +<br/>shape constraint during detection. A combination of a two-level boosted regression, shape +<br/>performance. In this paper, we have explain the advantage of ESR for deformable object like +<br/>face pose estimation and reveal its generic applications of the method. In the experiment, +<br/>we compare the results with different work and demonstrate the accuracy and robustness in +<br/>different scenarios. +<br/>Introduction +<br/>Pose estimation is an important problem in computer vision, and has enabled many practical ap- +<br/>plication from face expression 1 to activity tracking 2. Researchers design a new algorithm called +<br/>explicit shape regression (ESR) to find out face alignment from a picture 3. Figure 1 shows how +<br/>the system uses ESR to learn a shape of a human face image. A simple way to identify a face is to +<br/>find out facial landmarks like eyes, nose, mouth and chin. The researchers define a face shape S +<br/>and S is composed of Nf p facial landmarks. Therefore, they get S = [x1, y1, ..., xNf p, yNf p]T . The +<br/>objective of the researchers is to estimate a shape S of a face image. The way to know the accuracy +</td></tr><tr><td>1fc249ec69b3e23856b42a4e591c59ac60d77118</td><td>Evaluation of a 3D-aided Pose Invariant 2D Face Recognition System +<br/>Computational Biomedicine Lab +<br/>4800 Calhoun Rd. Houston, TX, USA +</td></tr><tr><td>1fbde67e87890e5d45864e66edb86136fbdbe20e</td><td>The Action Similarity Labeling Challenge +</td></tr><tr><td>1f41a96589c5b5cee4a55fc7c2ce33e1854b09d6</td><td>Demographic Estimation from Face Images: +<br/>Human vs. Machine Performance +</td></tr><tr><td>1fd2ed45fb3ba77f10c83f0eef3b66955645dfe0</td><td></td></tr><tr><td>1f2d12531a1421bafafe71b3ad53cb080917b1a7</td><td></td></tr><tr><td>1fefb2f8dd1efcdb57d5c2966d81f9ab22c1c58d</td><td>vExplorer: A Search Method to Find Relevant YouTube Videos for Health +<br/>Researchers +<br/>IBM Research, Cambridge, MA, USA +</td></tr><tr><td>1f94734847c15fa1da68d4222973950d6b683c9e</td><td>Embedding Label Structures for Fine-Grained Feature Representation +<br/>UNC Charlotte +<br/>Charlotte, NC 28223 +<br/>NEC Lab America +<br/>Cupertino, CA 95014 +<br/>NEC Lab America +<br/>Cupertino, CA 95014 +<br/>UNC Charlotte +<br/>Charlotte, NC 28223 +</td></tr><tr><td>1f745215cda3a9f00a65166bd744e4ec35644b02</td><td>Facial Cosmetics Database and Impact Analysis on +<br/>Automatic Face Recognition +<br/># Computer Science Department, TU Muenchen +<br/>Boltzmannstr. 3, 85748 Garching b. Muenchen, Germany +<br/>∗ Multimedia Communications Department, EURECOM +<br/>450 Route des Chappes, 06410 Biot, France +</td></tr><tr><td>1fff309330f85146134e49e0022ac61ac60506a9</td><td>Data-Driven Sparse Sensor Placement for Reconstruction +</td></tr><tr><td>7323b594d3a8508f809e276aa2d224c4e7ec5a80</td><td>JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 +<br/>An Experimental Evaluation of Covariates +<br/>Effects on Unconstrained Face Verification +</td></tr><tr><td>732e8d8f5717f8802426e1b9debc18a8361c1782</td><td>Unimodal Probability Distributions for Deep Ordinal Classification +</td></tr><tr><td>73ed64803d6f2c49f01cffef8e6be8fc9b5273b8</td><td>Noname manuscript No. +<br/>(will be inserted by the editor) +<br/>Cooking in the kitchen: Recognizing and Segmenting Human +<br/>Activities in Videos +<br/>Received: date / Accepted: date +</td></tr><tr><td>7306d42ca158d40436cc5167e651d7ebfa6b89c1</td><td>Noname manuscript No. +<br/>(will be inserted by the editor) +<br/>Transductive Zero-Shot Action Recognition by +<br/>Word-Vector Embedding +<br/>Received: date / Accepted: date +</td></tr><tr><td>734cdda4a4de2a635404e4c6b61f1b2edb3f501d</td><td>Tie and Guan EURASIP Journal on Image and Video Processing 2013, 2013:8 +<br/>http://jivp.eurasipjournals.com/content/2013/1/8 +<br/>R ES EAR CH +<br/>Open Access +<br/>Automatic landmark point detection and tracking +<br/>for human facial expressions +</td></tr><tr><td>732686d799d760ccca8ad47b49a8308b1ab381fb</td><td>Running head: TEACHERS’ DIFFERING BEHAVIORS +<br/>1 +<br/>Graduate School of Psychology +<br/>RESEARCH MASTER’S PSYCHOLOGY THESıS REPORT +<br/> +<br/>Teachers’ differing classroom behaviors: +<br/>The role of emotional sensitivity and cultural tolerance +<br/>Research Master’s, Social Psychology +<br/>Ethics Committee Reference Code: 2016-SP-7084 +</td></tr><tr><td>73fbdd57270b9f91f2e24989178e264f2d2eb7ae</td><td>978-1-4673-0046-9/12/$26.00 ©2012 IEEE +<br/>1945 +<br/>ICASSP 2012 +</td></tr><tr><td>73c9cbbf3f9cea1bc7dce98fce429bf0616a1a8c</td><td></td></tr><tr><td>871f5f1114949e3ddb1bca0982086cc806ce84a8</td><td>Discriminative Learning of Apparel Features +<br/>1 Computer Vision Laboratory, D-ITET, ETH Z¨urich, Switzerland +<br/>2 ESAT - PSI / IBBT, K.U. Leuven, Belgium +</td></tr><tr><td>878169be6e2c87df2d8a1266e9e37de63b524ae7</td><td>CBMM Memo No. 089 +<br/>May 10, 2018 +<br/>Image interpretation above and below the object level +</td></tr><tr><td>878301453e3d5cb1a1f7828002ea00f59cbeab06</td><td>Faceness-Net: Face Detection through +<br/>Deep Facial Part Responses +</td></tr><tr><td>87e592ee1a7e2d34e6b115da08700a1ae02e9355</td><td>Deep Pictorial Gaze Estimation +<br/>AIT Lab, Department of Computer Science, ETH Zurich +</td></tr><tr><td>87bb183d8be0c2b4cfceb9ee158fee4bbf3e19fd</td><td>Craniofacial Image Analysis +</td></tr><tr><td>8006219efb6ab76754616b0e8b7778dcfb46603d</td><td>CONTRIBUTIONSTOLARGE-SCALELEARNINGFORIMAGECLASSIFICATIONZeynepAkataPhDThesisl’´EcoleDoctoraleMath´ematiques,SciencesetTechnologiesdel’Information,InformatiquedeGrenoble </td></tr><tr><td>80193dd633513c2d756c3f568ffa0ebc1bb5213e</td><td></td></tr><tr><td>804b4c1b553d9d7bae70d55bf8767c603c1a09e3</td><td>978-1-4799-9988-0/16/$31.00 ©2016 IEEE +<br/>1831 +<br/>ICASSP 2016 +</td></tr><tr><td>800cbbe16be0f7cb921842d54967c9a94eaa2a65</td><td>MULTIMODAL RECOGNITION OF +<br/>EMOTIONS +</td></tr><tr><td>803c92a3f0815dbf97e30c4ee9450fd005586e1a</td><td>Max-Mahalanobis Linear Discriminant Analysis Networks +</td></tr><tr><td>80345fbb6bb6bcc5ab1a7adcc7979a0262b8a923</td><td>Research Article +<br/>Soft Biometrics for a Socially Assistive Robotic +<br/>Platform +<br/>Open Access +</td></tr><tr><td>80a6bb337b8fdc17bffb8038f3b1467d01204375</td><td>Proceedings of the International Conference on Computer and Information Science and Technology +<br/>Ottawa, Ontario, Canada, May 11 – 12, 2015 +<br/>Paper No. 126 +<br/>Subspace LDA Methods for Solving the Small Sample Size +<br/>Problem in Face Recognition +<br/><b></b><br/>101 KwanFu Rd., Sec. 2, Hsinchu, Taiwan +</td></tr><tr><td>80097a879fceff2a9a955bf7613b0d3bfa68dc23</td><td>Active Self-Paced Learning for Cost-Effective and +<br/>Progressive Face Identification +</td></tr><tr><td>74408cfd748ad5553cba8ab64e5f83da14875ae8</td><td>Facial Expressions Tracking and Recognition: Database Protocols for Systems Validation +<br/>and Evaluation +</td></tr><tr><td>747d5fe667519acea1bee3df5cf94d9d6f874f20</td><td></td></tr><tr><td>74dbe6e0486e417a108923295c80551b6d759dbe</td><td>International Journal of Computer Applications (0975 – 8887) +<br/>Volume 45– No.11, May 2012 +<br/>An HMM based Model for Prediction of Emotional +<br/>Composition of a Facial Expression using both +<br/>Significant and Insignificant Action Units and +<br/>Associated Gender Differences +<br/>Department of Management and Information +<br/>Department of Management and Information +<br/>Systems Science +<br/>1603-1 Kamitomioka, Nagaoka +<br/>Niigata, Japan +<br/>Systems Science +<br/>1603-1 Kamitomioka, Nagaoka +<br/>Niigata, Japan +</td></tr><tr><td>747c25bff37b96def96dc039cc13f8a7f42dbbc7</td><td>EmoNets: Multimodal deep learning approaches for emotion +<br/>recognition in video +</td></tr><tr><td>74b0095944c6e29837c208307a67116ebe1231c8</td><td></td></tr><tr><td>74156a11c2997517061df5629be78428e1f09cbd</td><td>Cancún Center, Cancún, México, December 4-8, 2016 +<br/>978-1-5090-4846-5/16/$31.00 ©2016 IEEE +<br/>2784 +</td></tr><tr><td>745b42050a68a294e9300228e09b5748d2d20b81</td><td></td></tr><tr><td>749d605dd12a4af58de1fae6f5ef5e65eb06540e</td><td>Multi-Task Video Captioning with Video and Entailment Generation +<br/>UNC Chapel Hill +</td></tr><tr><td>74c19438c78a136677a7cb9004c53684a4ae56ff</td><td>RESOUND: Towards Action Recognition +<br/>without Representation Bias +<br/>UC San Diego +</td></tr><tr><td>7480d8739eb7ab97c12c14e75658e5444b852e9f</td><td>NEGREL ET AL.: REVISITED MLBOOST FOR FACE RETRIEVAL +<br/>MLBoost Revisited: A Faster Metric +<br/>Learning Algorithm for Identity-Based Face +<br/>Retrieval +<br/>Frederic Jurie +<br/>Normandie Univ, UNICAEN, +<br/>ENSICAEN, CNRS +<br/>France +</td></tr><tr><td>74ba4ab407b90592ffdf884a20e10006d2223015</td><td>Partial Face Detection in the Mobile Domain +</td></tr><tr><td>7405ed035d1a4b9787b78e5566340a98fe4b63a0</td><td>Self-Expressive Decompositions for +<br/>Matrix Approximation and Clustering +</td></tr><tr><td>744db9bd550bf5e109d44c2edabffec28c867b91</td><td>FX e-Makeup for Muscle Based Interaction +<br/>1 Department of Informatics, PUC-Rio, Rio de Janeiro, Brazil +<br/>2 Department of Mechanical Engineering, PUC-Rio, Rio de Janeiro, Brazil +<br/>3 Department of Administration, PUC-Rio, Rio de Janeiro, Brazil +</td></tr><tr><td>744d23991a2c48d146781405e299e9b3cc14b731</td><td>This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIP.2016.2535284, IEEE +<br/>Transactions on Image Processing +<br/>Aging Face Recognition: A Hierarchical Learning +<br/>Model Based on Local Patterns Selection +</td></tr><tr><td>1a45ddaf43bcd49d261abb4a27977a952b5fff12</td><td>LDOP: Local Directional Order Pattern for Robust +<br/>Face Retrieval +<br/> +</td></tr><tr><td>1aa766bbd49bac8484e2545c20788d0f86e73ec2</td><td> +<br/>Baseline Face Detection, Head Pose Estimation, and Coarse +<br/>Direction Detection for Facial Data in the SHRP2 Naturalistic +<br/>Driving Study +<br/>J. Paone, D. Bolme, R. Ferrell, Member, IEEE, D. Aykac, and +<br/>T. Karnowski, Member, IEEE +<br/>Oak Ridge National Laboratory, Oak Ridge, TN +</td></tr><tr><td>1a849b694f2d68c3536ed849ed78c82e979d64d5</td><td>This is a repository copy of Symmetric Shape Morphing for 3D Face and Head Modelling. +<br/>White Rose Research Online URL for this paper: +<br/>http://eprints.whiterose.ac.uk/131760/ +<br/>Version: Accepted Version +<br/>Proceedings Paper: +<br/>Dai, Hang, Pears, Nicholas Edwin orcid.org/0000-0001-9513-5634, Smith, William Alfred +<br/>Peter orcid.org/0000-0002-6047-0413 et al. (1 more author) (2018) Symmetric Shape +<br/>Morphing for 3D Face and Head Modelling. In: The 13th IEEE Conference on Automatic +<br/>Face and Gesture Recognition. IEEE . +<br/>Reuse +<br/>Items deposited in White Rose Research Online are protected by copyright, with all rights reserved unless +<br/>indicated otherwise. They may be downloaded and/or printed for private study, or other acts as permitted by +<br/>national copyright laws. The publisher or other rights holders may allow further reproduction and re-use of +<br/>the full text version. This is indicated by the licence information on the White Rose Research Online record +<br/>for the item. +<br/>Takedown +<br/>If you consider content in White Rose Research Online to be in breach of UK law, please notify us by +<br/>https://eprints.whiterose.ac.uk/ +</td></tr><tr><td>1a3eee980a2252bb092666cf15dd1301fa84860e</td><td>PCA GAUSSIANIZATION FOR IMAGE PROCESSING +<br/>Image Processing Laboratory (IPL), Universitat de Val`encia +<br/>Catedr´atico A. Escardino - 46980 Paterna, Val`encia, Spain +</td></tr><tr><td>1a031378cf1d2b9088a200d9715d87db8a1bf041</td><td>Workshop track - ICLR 2018 +<br/>DEEP DICTIONARY LEARNING: SYNERGIZING RE- +<br/>CONSTRUCTION AND CLASSIFICATION +</td></tr><tr><td>1a9337d70a87d0e30966ecd1d7a9b0bbc7be161f</td><td></td></tr><tr><td>1a9a192b700c080c7887e5862c1ec578012f9ed1</td><td>IEEE TRANSACTIONS ON SYSTEM, MAN AND CYBERNETICS, PART B +<br/>Discriminant Subspace Analysis for Face +<br/>Recognition with Small Number of Training +<br/>Samples +</td></tr><tr><td>1a8ccc23ed73db64748e31c61c69fe23c48a2bb1</td><td>Extensive Facial Landmark Localization +<br/>with Coarse-to-fine Convolutional Network Cascade +<br/>Megvii Inc. +</td></tr><tr><td>1ad97cce5fa8e9c2e001f53f6f3202bddcefba22</td><td>Grassmann Averages for Scalable Robust PCA +<br/>DIKU and MPIs T¨ubingen∗ +<br/>Denmark and Germany +<br/>DTU Compute∗ +<br/>Lyngby, Denmark +</td></tr><tr><td>1a1118cd4339553ad0544a0a131512aee50cf7de</td><td></td></tr><tr><td>1a7a2221fed183b6431e29a014539e45d95f0804</td><td>Person Identification Using Text and Image Data +<br/>David S. Bolme, J. Ross Beveridge and Adele E. Howe +<br/>Computer Science Department +<br/>Colorado State Univeristy +<br/>Fort Collins, Colorado 80523 +</td></tr><tr><td>28e0ed749ebe7eb778cb13853c1456cb6817a166</td><td></td></tr><tr><td>28b9d92baea72ec665c54d9d32743cf7bc0912a7</td><td></td></tr><tr><td>28d7029cfb73bcb4ad1997f3779c183972a406b4</td><td>Discriminative Nonlinear Analysis Operator +<br/>Learning: When Cosparse Model Meets Image +<br/>Classification +</td></tr><tr><td>280d59fa99ead5929ebcde85407bba34b1fcfb59</td><td>978-1-4799-9988-0/16/$31.00 ©2016 IEEE +<br/>2662 +<br/>ICASSP 2016 +</td></tr><tr><td>28cd46a078e8fad370b1aba34762a874374513a5</td><td>CVPAPER.CHALLENGE IN 2016, JULY 2017 +<br/>cvpaper.challenge in 2016: Futuristic Computer +<br/>Vision through 1,600 Papers Survey +</td></tr><tr><td>282a3ee79a08486f0619caf0ada210f5c3572367</td><td></td></tr><tr><td>288dbc40c027af002298b38954d648fddd4e2fd3</td><td></td></tr><tr><td>28312c3a47c1be3a67365700744d3d6665b86f22</td><td></td></tr><tr><td>28b5b5f20ad584e560cd9fb4d81b0a22279b2e7b</td><td>A New Fuzzy Stacked Generalization Technique +<br/>and Analysis of its Performance +</td></tr><tr><td>28bc378a6b76142df8762cd3f80f737ca2b79208</td><td>Understanding Objects in Detail with Fine-grained Attributes +<br/>Ross Girshick5 +<br/>David Weiss7 +</td></tr><tr><td>287900f41dd880802aa57f602e4094a8a9e5ae56</td><td></td></tr><tr><td>28d4e027c7e90b51b7d8908fce68128d1964668a</td><td></td></tr><tr><td>2866cbeb25551257683cf28f33d829932be651fe</td><td>In Proceedings of the 2018 IEEE International Conference on Image Processing (ICIP) +<br/>The final publication is available at: http://dx.doi.org/10.1109/ICIP.2018.8451026 +<br/>A TWO-STEP LEARNING METHOD FOR DETECTING LANDMARKS +<br/>ON FACES FROM DIFFERENT DOMAINS +<br/>Erickson R. Nascimento +<br/>Universidade Federal de Minas Gerais (UFMG), Brazil +</td></tr><tr><td>28aa89b2c827e5dd65969a5930a0520fdd4a3dc7</td><td></td></tr><tr><td>28b061b5c7f88f48ca5839bc8f1c1bdb1e6adc68</td><td>Predicting User Annoyance Using Visual Attributes +<br/>Virginia Tech +<br/>Goibibo +<br/>Virginia Tech +<br/>Virginia Tech +</td></tr><tr><td>17a85799c59c13f07d4b4d7cf9d7c7986475d01c</td><td>ADVERTIMENT. La consulta d’aquesta tesi queda condicionada a l’acceptació de les següents +<br/>condicions d'ús: La difusió d’aquesta tesi per mitjà del servei TDX (www.tesisenxarxa.net) ha +<br/>estat autoritzada pels titulars dels drets de propietat intel·lectual únicament per a usos privats +<br/>emmarcats en activitats d’investigació i docència. No s’autoritza la seva reproducció amb finalitats +<br/>de lucre ni la seva difusió i posada a disposició des d’un lloc aliè al servei TDX. No s’autoritza la +<br/>presentació del seu contingut en una finestra o marc aliè a TDX (framing). Aquesta reserva de +<br/>drets afecta tant al resum de presentació de la tesi com als seus continguts. En la utilització o cita +<br/>de parts de la tesi és obligat indicar el nom de la persona autora. +<br/>ADVERTENCIA. La consulta de esta tesis queda condicionada a la aceptación de las siguientes +<br/>condiciones de uso: La difusión de esta tesis por medio del servicio TDR (www.tesisenred.net) ha +<br/>sido autorizada por los titulares de los derechos de propiedad intelectual únicamente para usos +<br/>privados enmarcados en actividades de investigación y docencia. No se autoriza su reproducción +<br/>con finalidades de lucro ni su difusión y puesta a disposición desde un sitio ajeno al servicio TDR. +<br/>No se autoriza la presentación de su contenido en una ventana o marco ajeno a TDR (framing). +<br/>Esta reserva de derechos afecta tanto al resumen de presentación de la tesis como a sus +<br/>contenidos. En la utilización o cita de partes de la tesis es obligado indicar el nombre de la +<br/>persona autora. +<br/>WARNING. On having consulted this thesis you’re accepting the following use conditions: +<br/>Spreading this thesis by the TDX (www.tesisenxarxa.net) service has been authorized by the +<br/>titular of the intellectual property rights only for private uses placed in investigation and teaching +<br/>activities. Reproduction with lucrative aims is not authorized neither its spreading and availability +<br/>from a site foreign to the TDX service. Introducing its content in a window or frame foreign to the +<br/>TDX service is not authorized (framing). This rights affect to the presentation summary of the +<br/>thesis as well as to its contents. In the using or citation of parts of the thesis it’s obliged to indicate +<br/>the name of the author +</td></tr><tr><td>176f26a6a8e04567ea71677b99e9818f8a8819d0</td><td>MEG: Multi-Expert Gender classification from +<br/>face images in a demographics-balanced dataset +</td></tr><tr><td>17035089959a14fe644ab1d3b160586c67327db2</td><td></td></tr><tr><td>17a995680482183f3463d2e01dd4c113ebb31608</td><td>IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. X, NO. Y, MONTH Z +<br/>Structured Label Inference for +<br/>Visual Understanding +</td></tr><tr><td>17aa78bd4331ef490f24bdd4d4cd21d22a18c09c</td><td></td></tr><tr><td>17c0d99171efc957b88c31a465c59485ab033234</td><td></td></tr><tr><td>1742ffea0e1051b37f22773613f10f69d2e4ed2c</td><td></td></tr><tr><td>1791f790b99471fc48b7e9ec361dc505955ea8b1</td><td></td></tr><tr><td>174930cac7174257515a189cd3ecfdd80ee7dd54</td><td>Multi-view Face Detection Using Deep Convolutional +<br/>Neural Networks +<br/>Yahoo +<br/>Mohammad Saberian +<br/>inc.com +<br/>Yahoo +<br/>Yahoo +</td></tr><tr><td>17fad2cc826d2223e882c9fda0715fcd5475acf3</td><td></td></tr><tr><td>1750db78b7394b8fb6f6f949d68f7c24d28d934f</td><td>Detecting Facial Retouching Using Supervised +<br/>Deep Learning +<br/>Bowyer, Fellow, IEEE +</td></tr><tr><td>173657da03e3249f4e47457d360ab83b3cefbe63</td><td>HKU-Face: A Large Scale Dataset for +<br/>Deep Face Recognition +<br/>Final Report +<br/>3035140108 +<br/>COMP4801 Final Year Project +<br/>Project Code: 17007 +</td></tr><tr><td>7ba0bf9323c2d79300f1a433ff8b4fe0a00ad889</td><td></td></tr><tr><td>7bfe085c10761f5b0cc7f907bdafe1ff577223e0</td><td></td></tr><tr><td>7b9b3794f79f87ca8a048d86954e0a72a5f97758</td><td>DOI 10.1515/jisys-2013-0016 Journal of Intelligent Systems 2013; 22(4): 365–415 +<br/>Passing an Enhanced Turing Test – +<br/>Interacting with Lifelike Computer +<br/>Representations of Specific Individuals +</td></tr><tr><td>7b0f1fc93fb24630eb598330e13f7b839fb46cce</td><td>Learning to Find Eye Region Landmarks for Remote Gaze +<br/>Estimation in Unconstrained Settings +<br/>ETH Zurich +<br/>MPI for Informatics +<br/>MPI for Informatics +<br/>ETH Zurich +</td></tr><tr><td>7bdcd85efd1e3ce14b7934ff642b76f017419751</td><td>289 +<br/>Learning Discriminant Face Descriptor +</td></tr><tr><td>7b3b7769c3ccbdf7c7e2c73db13a4d32bf93d21f</td><td>On the Design and Evaluation of Robust Head Pose for +<br/>Visual User Interfaces: Algorithms, Databases, and +<br/>Comparisons +<br/>Laboratory of Intelligent and +<br/>Safe Automobiles +<br/>UCSD - La Jolla, CA, USA +<br/>Laboratory of Intelligent and +<br/>Safe Automobiles +<br/>UCSD - La Jolla, CA, USA +<br/>Laboratory of Intelligent and +<br/>Safe Automobiles +<br/>UCSD - La Jolla, CA, USA +<br/>Laboratory of Intelligent and +<br/>Safe Automobiles +<br/>UCSD - La Jolla, CA, USA +<br/>Mohan Trivedi +<br/>Laboratory of Intelligent and +<br/>Safe Automobiles +<br/>UCSD - La Jolla, CA, USA +</td></tr><tr><td>8f772d9ce324b2ef5857d6e0b2a420bc93961196</td><td>MAHPOD et al.: CFDRNN +<br/>Facial Landmark Point Localization using +<br/>Coarse-to-Fine Deep Recurrent Neural Network +</td></tr><tr><td>8fb611aca3bd8a3a0527ac0f38561a5a9a5b8483</td><td></td></tr><tr><td>8fda2f6b85c7e34d3e23927e501a4b4f7fc15b2a</td><td>Feature Selection with Annealing for Big Data +<br/>Learning +</td></tr><tr><td>8f9c37f351a91ed416baa8b6cdb4022b231b9085</td><td>Generative Adversarial Style Transfer Networks for Face Aging +<br/>Sveinn Palsson +<br/>D-ITET, ETH Zurich +<br/>Eirikur Agustsson +<br/>D-ITET, ETH Zurich +</td></tr><tr><td>8f8c0243816f16a21dea1c20b5c81bc223088594</td><td></td></tr><tr><td>8f89aed13cb3555b56fccd715753f9ea72f27f05</td><td>Attended End-to-end Architecture for Age +<br/>Estimation from Facial Expression Videos +</td></tr><tr><td>8f9f599c05a844206b1bd4947d0524234940803d</td><td></td></tr><tr><td>8fd9c22b00bd8c0bcdbd182e17694046f245335f</td><td> +<br/>Recognizing Facial Expressions in Videos +</td></tr><tr><td>8a866bc0d925dfd8bb10769b8b87d7d0ff01774d</td><td>WikiArt Emotions: An Annotated Dataset of Emotions Evoked by Art +<br/>National Research Council Canada +</td></tr><tr><td>8a40b6c75dd6392ee0d3af73cdfc46f59337efa9</td><td></td></tr><tr><td>8a91ad8c46ca8f4310a442d99b98c80fb8f7625f</td><td>2592 +<br/>2D Segmentation Using a Robust Active +<br/>Shape Model With the EM Algorithm +</td></tr><tr><td>8aed6ec62cfccb4dba0c19ee000e6334ec585d70</td><td>Localizing and Visualizing Relative Attributes +</td></tr><tr><td>8a336e9a4c42384d4c505c53fb8628a040f2468e</td><td>Wang and Luo EURASIP Journal on Bioinformatics +<br/>and Systems Biology (2016) 2016:13 +<br/>DOI 10.1186/s13637-016-0048-7 +<br/>R ES EAR CH +<br/>Detecting Visually Observable Disease +<br/>Symptoms from Faces +<br/>Open Access +</td></tr><tr><td>7e600faee0ba11467d3f7aed57258b0db0448a72</td><td></td></tr><tr><td>7e8016bef2c180238f00eecc6a50eac473f3f138</td><td>TECHNISCHE UNIVERSIT ¨AT M ¨UNCHEN +<br/>Lehrstuhl f¨ur Mensch-Maschine-Kommunikation +<br/>Immersive Interactive Data Mining and Machine +<br/>Learning Algorithms for Big Data Visualization +<br/>Vollst¨andiger Abdruck der von der Fakult¨at f¨ur Elektrotechnik und Informationstechnik +<br/>der Technischen Universit¨at M¨unchen zur Erlangung des akademischen Grades eines +<br/>Doktor-Ingenieurs (Dr.-Ing.) +<br/>genehmigten Dissertation. +<br/>Vorsitzender: +<br/>Univ.-Prof. Dr. sc.techn. Andreas Herkersdorf +<br/>Pr¨ufer der Dissertation: +<br/>1. Univ.-Prof. Dr.-Ing. habil. Gerhard Rigoll +<br/>2. Univ.-Prof. Dr.-Ing. habil. Dirk Wollherr +<br/>3. Prof. Dr. Mihai Datcu +<br/>Die Dissertation wurde am 13.08.2015 bei der Technischen Universit¨at M¨unchen eingerei- +<br/>cht und durch die Fakult¨at f¨ur Elektrotechnik und Informationstechnik am 16.02.2016 +<br/>angenommen. +</td></tr><tr><td>7e3367b9b97f291835cfd0385f45c75ff84f4dc5</td><td>Improved Local Binary Pattern Based Action Unit Detection Using +<br/>Morphological and Bilateral Filters +<br/>1Signal Processing Laboratory (LTS5) +<br/>´Ecole Polytechnique F´ed´erale de Lausanne, +<br/>Switzerland +<br/>2nViso SA +<br/>Lausanne, Switzerland +</td></tr><tr><td>7ed6ff077422f156932fde320e6b3bd66f8ffbcb</td><td>State of 3D Face Biometrics for Homeland Security Applications +<br/>Chaudhari4 +</td></tr><tr><td>7e507370124a2ac66fb7a228d75be032ddd083cc</td><td>This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAFFC.2017.2708106, IEEE +<br/>Transactions on Affective Computing +<br/>Dynamic Pose-Robust Facial Expression +<br/>Recognition by Multi-View Pairwise Conditional +<br/>Random Forests +<br/>1 Sorbonne Universit´es, UPMC Univ Paris 06 +<br/>CNRS, UMR 7222, F-75005, Paris, France +</td></tr><tr><td>1056347fc5e8cd86c875a2747b5f84fd570ba232</td><td></td></tr><tr><td>10e7dd3bbbfbc25661213155e0de1a9f043461a2</td><td>Cross Euclidean-to-Riemannian Metric Learning +<br/>with Application to Face Recognition from Video +</td></tr><tr><td>10ab1b48b2a55ec9e2920a5397febd84906a7769</td><td></td></tr><tr><td>10ce3a4724557d47df8f768670bfdd5cd5738f95</td><td>Fihe igh Fie +<br/>Ac e ad +<br/>Ra +<br/>The Rbic i e Caegie e +<br/>5000 Fbe Ave e ib gh A 15213 +<br/>Abac. ay face ecgii ak he e ad i +<br/>cdii f he be ad ga +<br/> +<br/>a di(cid:11)ee e ad de a di(cid:11)ee i +<br/>ecgii a +<br/> bjec ca ed a abiay e ad de abiay i +<br/>ad ay be f be iage agai ca ed a abiay e ad +<br/> de abiay i +<br/>Fihe +<br/>iage. achig bewee he be ad ga +<br/>he Fihe +<br/>d ci +<br/> ay face ecgii ceai he e f he be ad ga +<br/>di(cid:11)ee. The ga +<br/>The a +<br/>ga +<br/>view ca ed f a caea i he ce f he . The be f ga +<br/>ad be iage ca a +<br/>iage f each bjec a fa +<br/>yica +<br/>iage a ig +<br/>Face ecgii ac e i.e. face ecgii whee he ga +<br/>iage d have he ae e ha eceived vey +<br/>have bee ed which ca ecgize face [1] e geea +<br/>a a vaiey f e. +<br/>a evey e. A +<br/>f exa +<br/>caiig a +<br/>iai vaiai. ca be ed wih abiay ga +<br/>Afe e vaiai he ex igi(cid:12)ca fac a(cid:11)ecig he aea +<br/>ace f face i i +<br/>face ecgii ac i +<br/>face [4 5]. +<br/>i +</td></tr><tr><td>102e374347698fe5404e1d83f441630b1abf62d9</td><td>Facial Image Analysis for Fully-Automatic +<br/>Prediction of Difficult Endotracheal Intubation +</td></tr><tr><td>100641ed8a5472536dde53c1f50fa2dd2d4e9be9</td><td>Visual Attributes for Enhanced Human-Machine Communication* +</td></tr><tr><td>10195a163ab6348eef37213a46f60a3d87f289c5</td><td></td></tr><tr><td>10e704c82616fb5d9c48e0e68ee86d4f83789d96</td><td></td></tr><tr><td>101569eeef2cecc576578bd6500f1c2dcc0274e2</td><td>Multiaccuracy: Black-Box Post-Processing for Fairness in +<br/>Classification +<br/>James Zou +</td></tr><tr><td>106732a010b1baf13c61d0994552aee8336f8c85</td><td>Expanded Parts Model for Semantic Description +<br/>of Humans in Still Images +</td></tr><tr><td>10e70a34d56258d10f468f8252a7762950830d2b</td><td></td></tr><tr><td>102b27922e9bd56667303f986404f0e1243b68ab</td><td>Wang et al. Appl Inform (2017) 4:13 +<br/>DOI 10.1186/s40535-017-0042-5 +<br/>RESEARCH +<br/>Multiscale recurrent regression networks +<br/>for face alignment +<br/>Open Access +<br/>*Correspondence: +<br/>3 State Key Lab of Intelligent +<br/>Technologies and Systems, +<br/>Beijing 100084, People’s +<br/>Republic of China +<br/>Full list of author information +<br/>is available at the end of the +<br/>article +</td></tr><tr><td>10fcbf30723033a5046db791fec2d3d286e34daa</td><td>On-Line Cursive Handwriting Recognition: A Survey of Methods +<br/>and Performances +<br/>*Faculty of Computer Science & Information Systems, Universiti Teknologi Malaysia (UTM) , 81310 +<br/>Skudai, Johor, Malaysia. +</td></tr><tr><td>108b2581e07c6b7ca235717c749d45a1fa15bb24</td><td>Using Stereo Matching with General Epipolar +<br/>Geometry for 2D Face Recognition +<br/>across Pose +</td></tr><tr><td>10d334a98c1e2a9e96c6c3713aadd42a557abb8b</td><td>Scene Text Recognition using Part-based Tree-structured Character Detection +<br/>State Key Laboratory of Management and Control for Complex Systems, CASIA, Beijing, China +</td></tr><tr><td>192723085945c1d44bdd47e516c716169c06b7c0</td><td>This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation +<br/>Vision and Attention Theory Based Sampling +<br/>for Continuous Facial Emotion Recognition +<br/>Ninad S. Thakoor, Member, IEEE +<br/>10 +<br/>11 +<br/>12 +<br/>13 +<br/>14 +<br/>15 +<br/>16 +<br/>17 +<br/>18 +<br/>19 +<br/>20 +<br/>21 +<br/>22 +<br/>23 +<br/>24 +<br/>25 +<br/>26 +<br/>27 +<br/>28 +<br/>29 +<br/>30 +<br/>31 +<br/>32 +<br/>33 +<br/>34 +<br/>35 +<br/>36 +<br/>37 +</td></tr><tr><td>19fb5e5207b4a964e5ab50d421e2549ce472baa8</td><td>International Conference on Computer Systems and Technologies - CompSysTech’14 +<br/>Online Emotional Facial Expression Dictionary +<br/>Léon Rothkrantz +</td></tr><tr><td>1962e4c9f60864b96c49d85eb897141486e9f6d1</td><td>Neural Comput & Applic (2011) 20:565–573 +<br/>DOI 10.1007/s00521-011-0577-7 +<br/>O R I G I N A L A R T I C L E +<br/>Locality preserving embedding for face and handwriting digital +<br/>recognition +<br/>Received: 3 December 2008 / Accepted: 11 March 2011 / Published online: 1 April 2011 +<br/>Ó Springer-Verlag London Limited 2011 +<br/>supervised manifold +<br/>the local sub-manifolds. +</td></tr><tr><td>19af008599fb17bbd9b12288c44f310881df951c</td><td>Discriminative Local Sparse Representations for +<br/>Robust Face Recognition +</td></tr><tr><td>19296e129c70b332a8c0a67af8990f2f4d4f44d1</td><td>Metric Learning Approaches for Face Identification +<br/>Is that you? +<br/>M. Guillaumin, J. Verbeek and C. Schmid +<br/>LEAR team, INRIA Rhˆone-Alpes, France +<br/>Supplementary Material +</td></tr><tr><td>19666b9eefcbf764df7c1f5b6938031bcf777191</td><td>Group Component Analysis for Multi-block Data: +<br/>Common and Individual Feature Extraction +</td></tr><tr><td>190b3caa2e1a229aa68fd6b1a360afba6f50fde4</td><td></td></tr><tr><td>19c0c7835dba1a319b59359adaa738f0410263e8</td><td>228 +<br/>Natural Image Statistics and +<br/>Low-Complexity Feature Selection +</td></tr><tr><td>19808134b780b342e21f54b60095b181dfc7a600</td><td></td></tr><tr><td>19d583bf8c5533d1261ccdc068fdc3ef53b9ffb9</td><td>FaceNet: A Unified Embedding for Face Recognition and Clustering +<br/>Google Inc. +<br/>Google Inc. +<br/>Google Inc. +</td></tr><tr><td>197c64c36e8a9d624a05ee98b740d87f94b4040c</td><td>Regularized Greedy Column Subset Selection +<br/>aDepartment of Computer Systems, Universidad Polit´ecnica de Madrid +<br/>bDepartment of Applied Mathematics, Universidad Polit´ecnica de Madrid +</td></tr><tr><td>19d4855f064f0d53cb851e9342025bd8503922e2</td><td>Learning SURF Cascade for Fast and Accurate Object Detection +<br/>Intel Labs China +</td></tr><tr><td>19eb486dcfa1963c6404a9f146c378fc7ae3a1df</td><td></td></tr><tr><td>4c6daffd092d02574efbf746d086e6dc0d3b1e91</td><td></td></tr><tr><td>4c6e1840451e1f86af3ef1cb551259cb259493ba</td><td>HAND POSTURE DATASET CREATION FOR GESTURE +<br/>RECOGNITION +<br/>Instituto de Sistemas Inteligentes y Aplicaciones Numericas en Ingenieria +<br/>Campus Universitario de Tafira, 35017 Gran Canaria, Spain +<br/>Departamento de E.I.O. y Computacion +<br/>38271 Universidad de La Laguna, Spain +<br/>Keywords: +<br/>Image understanding, Gesture recognition, Hand dataset. +</td></tr><tr><td>4c29e1f31660ba33e46d7e4ffdebb9b8c6bd5adc</td><td></td></tr><tr><td>4c815f367213cc0fb8c61773cd04a5ca8be2c959</td><td>978-1-4244-4296-6/10/$25.00 ©2010 IEEE +<br/>2470 +<br/>ICASSP 2010 +</td></tr><tr><td>4c4236b62302957052f1bbfbd34dbf71ac1650ec</td><td>SEMI-SUPERVISED FACE RECOGNITION WITH LDA SELF-TRAINING +<br/>Multimedia Communications Department, EURECOM +<br/>2229 Route des Crêtes , BP 193, F-06560 Sophia-Antipolis Cedex, France +</td></tr><tr><td>2661f38aaa0ceb424c70a6258f7695c28b97238a</td><td>IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART B: CYBERNETICS, VOL. 42, NO. 4, AUGUST 2012 +<br/>1027 +<br/>Multilayer Architectures for Facial +<br/>Action Unit Recognition +</td></tr><tr><td>2609079d682998da2bc4315b55a29bafe4df414e</td><td>ON RANK AGGREGATION FOR FACE RECOGNITION FROM VIDEOS +<br/>IIIT-Delhi, India +</td></tr><tr><td>26a72e9dd444d2861298d9df9df9f7d147186bcd</td><td>DOI 10.1007/s00138-016-0768-4 +<br/>ORIGINAL PAPER +<br/>Collecting and annotating the large continuous action dataset +<br/>Received: 18 June 2015 / Revised: 18 April 2016 / Accepted: 22 April 2016 / Published online: 21 May 2016 +<br/>© The Author(s) 2016. This article is published with open access at Springerlink.com +</td></tr><tr><td>265af79627a3d7ccf64e9fe51c10e5268fee2aae</td><td>1817 +<br/>A Mixture of Transformed Hidden Markov +<br/>Models for Elastic Motion Estimation +</td></tr><tr><td>267c6e8af71bab68547d17966adfaab3b4711e6b</td><td></td></tr><tr><td>26a89701f4d41806ce8dbc8ca00d901b68442d45</td><td></td></tr><tr><td>26ad6ceb07a1dc265d405e47a36570cb69b2ace6</td><td>RESEARCH AND EXPLOR ATORY +<br/>DEVELOPMENT DEPARTMENT +<br/>REDD-2015-384 +<br/>Neural Correlates of Cross-Cultural +<br/>How to Improve the Training and Selection for +<br/>Military Personnel Involved in Cross-Cultural +<br/>Operating Under Grant #N00014-12-1-0629/113056 +<br/>Adaptation +<br/>September, 2015 +<br/>Interactions +<br/>Prepared for: +<br/>Office of Naval Research +</td></tr><tr><td>26e570049aaedcfa420fc8c7b761bc70a195657c</td><td>J Sign Process Syst +<br/>DOI 10.1007/s11265-017-1276-0 +<br/>Hybrid Facial Regions Extraction for Micro-expression +<br/>Recognition System +<br/>Received: 2 February 2016 / Revised: 20 October 2016 / Accepted: 10 August 2017 +<br/>© Springer Science+Business Media, LLC 2017 +</td></tr><tr><td>21ef129c063bad970b309a24a6a18cbcdfb3aff5</td><td>POUR L'OBTENTION DU GRADE DE DOCTEUR ÈS SCIENCESacceptée sur proposition du jury:Dr J.-M. Vesin, président du juryProf. J.-Ph. Thiran, Prof. D. Sander, directeurs de thèseProf. M. F. Valstar, rapporteurProf. H. K. Ekenel, rapporteurDr S. Marcel, rapporteurIndividual and Inter-related Action Unit Detection in Videos for Affect RecognitionTHÈSE NO 6837 (2016)ÉCOLE POLYTECHNIQUE FÉDÉRALE DE LAUSANNEPRÉSENTÉE LE 19 FÉVRIER 2016À LA FACULTÉ DES SCIENCES ET TECHNIQUES DE L'INGÉNIEURLABORATOIRE DE TRAITEMENT DES SIGNAUX 5PROGRAMME DOCTORAL EN GÉNIE ÉLECTRIQUE Suisse2016PARAnıl YÜCE </td></tr><tr><td>218b2c5c9d011eb4432be4728b54e39f366354c1</td><td>Enhancing Training Collections for Image +<br/>Annotation: An Instance-Weighted Mixture +<br/>Modeling Approach +</td></tr><tr><td>21e828071249d25e2edaca0596e27dcd63237346</td><td></td></tr><tr><td>2162654cb02bcd10794ae7e7d610c011ce0fb51b</td><td>4697 +<br/>978-1-4799-5751-4/14/$31.00 ©2014 IEEE +<br/>1http://www.skype.com/ +<br/>2http://www.google.com/hangouts/ +<br/>tification, sparse coding +</td></tr><tr><td>21f3c5b173503185c1e02a3eb4e76e13d7e9c5bc</td><td>m a s s a c h u s e t t s i n s t i t u t e o f +<br/>t e c h n o l o g y — a r t i f i c i a l i n t e l l i g e n c e l a b o r a t o r y +<br/>Rotation Invariant Real-time +<br/>Face Detection and +<br/>Recognition System +<br/>AI Memo 2001-010 +<br/>CBCL Memo 197 +<br/>May 31, 2001 +<br/>© 2 0 0 1 m a s s a c h u s e t t s i n s t i t u t e o f +<br/>t e c h n o l o g y, c a m b r i d g e , m a 0 2 1 3 9 u s a — w w w. a i . m i t . e d u +</td></tr><tr><td>21bd9374c211749104232db33f0f71eab4df35d5</td><td>Integrating Facial Makeup Detection Into +<br/>Multimodal Biometric User Verification System +<br/>CuteSafe Technology Inc. +<br/>Gebze, Kocaeli, Turkey +<br/>Eurecom Digital Security Department +<br/>06410 Biot, France +</td></tr><tr><td>213a579af9e4f57f071b884aa872651372b661fd</td><td>Int J Comput Vis +<br/>DOI 10.1007/s11263-013-0672-6 +<br/>Automatic and Efficient Human Pose Estimation for Sign +<br/>Language Videos +<br/>Received: 4 February 2013 / Accepted: 29 October 2013 +<br/>© Springer Science+Business Media New York 2013 +</td></tr><tr><td>21626caa46cbf2ae9e43dbc0c8e789b3dbb420f1</td><td>978-1-4673-2533-2/12/$26.00 ©2012 IEEE +<br/>1437 +<br/>ICIP 2012 +</td></tr><tr><td>4d49c6cff198cccb21f4fa35fd75cbe99cfcbf27</td><td>Topological Principal Component Analysis for +<br/>face encoding and recognition +<br/>Juan J. Villanueva +<br/>Computer Vision Center and Departament d’Inform(cid:18)atica, Edi(cid:12)ci O, Universitat +<br/>Aut(cid:18)onoma de Barcelona +</td></tr><tr><td>4da735d2ed0deeb0cae4a9d4394449275e316df2</td><td>Gothenburg, Sweden, June 19-22, 2016 +<br/>978-1-5090-1820-8/16/$31.00 ©2016 IEEE +<br/>1410 +</td></tr><tr><td>4d530a4629671939d9ded1f294b0183b56a513ef</td><td>International Journal of Machine Learning and Computing, Vol. 2, No. 4, August 2012 +<br/>Facial Expression Classification Method Based on Pseudo +<br/>Zernike Moment and Radial Basis Function Network +<br/> +</td></tr><tr><td>4d2975445007405f8cdcd74b7fd1dd547066f9b8</td><td>Image and Video Processing +<br/>for Affective Applications +</td></tr><tr><td>4df889b10a13021928007ef32dc3f38548e5ee56</td><td></td></tr><tr><td>4d423acc78273b75134e2afd1777ba6d3a398973</td><td></td></tr><tr><td>4db9e5f19366fe5d6a98ca43c1d113dac823a14d</td><td>Combining Crowdsourcing and Face Recognition to Identify Civil War Soldiers +<br/>Are 1,000 Features Worth A Picture? +<br/>Department of Computer Science and Center for Human-Computer Interaction +<br/>Virginia Tech, Arlington, VA, USA +</td></tr><tr><td>4dd6d511a8bbc4d9965d22d79ae6714ba48c8e41</td><td></td></tr><tr><td>4d7e1eb5d1afecb4e238ba05d4f7f487dff96c11</td><td>978-1-5090-4117-6/17/$31.00 ©2017 IEEE +<br/>2352 +<br/>ICASSP 2017 +</td></tr><tr><td>4d90bab42806d082e3d8729067122a35bbc15e8d</td><td></td></tr><tr><td>4d6ad0c7b3cf74adb0507dc886993e603c863e8c</td><td>Human Activity Recognition Based on Wearable +<br/>Sensor Data: A Standardization of the +<br/>State-of-the-Art +<br/>Smart Surveillance Interest Group, Computer Science Department +<br/>Universidade Federal de Minas Gerais, Brazil +</td></tr><tr><td>4d0ef449de476631a8d107c8ec225628a67c87f9</td><td>© 2010 IEEE. Personal use of this material is permitted. Permission from IEEE +<br/>must be obtained for all other uses, in any current or future media, including +<br/>reprinting/republishing this material for advertising or promotional purposes, +<br/>creating new collective works, for resale or redistribution to servers or lists, or +<br/>reuse of any copyrighted component of this work in other works. +<br/>Pre-print of article that appeared at BTAS 2010. +<br/>The published article can be accessed from: +<br/>http://ieeexplore.ieee.org/xpl/freeabs_all.jsp?arnumber=5634517 +</td></tr><tr><td>4d47261b2f52c361c09f7ab96fcb3f5c22cafb9f</td><td>Deep multi-frame face super-resolution +<br/>Evgeniya Ustinova, Victor Lempitsky +<br/>October 17, 2017 +</td></tr><tr><td>75879ab7a77318bbe506cb9df309d99205862f6c</td><td>Analysis Of Emotion Recognition From Facial +<br/>Expressions Using Spatial And Transform Domain +<br/>Methods +</td></tr><tr><td>7574f999d2325803f88c4915ba8f304cccc232d1</td><td>Transfer Learning For Cross-Dataset Recognition: A Survey +<br/>This paper summarises and analyses the cross-dataset recognition transfer learning techniques with the +<br/>emphasis on what kinds of methods can be used when the available source and target data are presented +<br/>in different forms for boosting the target task. This paper for the first time summarises several transferring +<br/>criteria in details from the concept level, which are the key bases to guide what kind of knowledge to transfer +<br/>between datasets. In addition, a taxonomy of cross-dataset scenarios and problems is proposed according the +<br/>properties of data that define how different datasets are diverged, thereby review the recent advances on +<br/>each specific problem under different scenarios. Moreover, some real world applications and corresponding +<br/>commonly used benchmarks of cross-dataset recognition are reviewed. Lastly, several future directions are +<br/>identified. +<br/>Additional Key Words and Phrases: Cross-dataset, transfer learning, domain adaptation +<br/>1. INTRODUCTION +<br/>It has been explored how human would transfer learning in one context to another +<br/>similar context [Woodworth and Thorndike 1901; Perkins et al. 1992] in the field of +<br/>Psychology and Education. For example, learning to drive a car helps a person later +<br/>to learn more quickly to drive a truck, and learning mathematics prepares students to +<br/>study physics. The machine learning algorithms are mostly inspired by human brains. +<br/>However, most of them require a huge amount of training examples to learn a new +<br/>model from scratch and fail to apply knowledge learned from previous domains or +<br/>tasks. This may be due to that a basic assumption of statistical learning theory is +<br/>that the training and test data are drawn from the same distribution and belong to +<br/>the same task. Intuitively, learning from scratch is not realistic and practical, because +<br/>it violates how human learn things. In addition, manually labelling a large amount +<br/>of data for new domain or task is labour extensive, especially for the modern “data- +<br/>hungry” and “data-driven” learning techniques (i.e. deep learning). However, the big +<br/>data era provides a huge amount available data collected for other domains and tasks. +<br/>Hence, how to use the previously available data smartly for the current task with +<br/>scarce data will be beneficial for real world applications. +<br/>To reuse the previous knowledge for current tasks, the differences between old data +<br/>and new data need to be taken into account. Take the object recognition as an ex- +<br/>ample. As claimed by Torralba and Efros [2011], despite the great efforts of object +<br/>datasets creators, the datasets appear to have strong build-in bias caused by various +<br/>factors, such as selection bias, capture bias, category or label bias, and negative set +<br/>bias. This suggests that no matter how big the dataset is, it is impossible to cover +<br/>the complexity of the real visual world. Hence, the dataset bias needs to be consid- +<br/>ered before reusing data from previous datasets. Pan and Yang [2010] summarise that +<br/>the differences between different datasets can be caused by domain divergence (i.e. +<br/>distribution shift or feature space difference) or task divergence (i.e. conditional dis- +<br/>tribution shift or label space difference), or both. For example, in visual recognition, +<br/>the distributions between the previous and current data can be discrepant due to the +<br/>different environments, lighting, background, sensor types, resolutions, view angles, +<br/>and post-processing. Those external factors may cause the distribution divergence or +<br/>even feature space divergence between different domains. On the other hand, the task +<br/>divergence between current and previous data is also ubiquitous. For example, it is +<br/>highly possible that an animal species that we want to recognize have not been seen +<br/>ACM Journal Name, Vol. V, No. N, Article A, Publication date: January YYYY. +</td></tr><tr><td>75e9a141b85d902224f849ea61ab135ae98e7bfb</td><td></td></tr><tr><td>75503aff70a61ff4810e85838a214be484a674ba</td><td>Improved Facial Expression Recognition via Uni-Hyperplane Classification +<br/>S.W. Chew∗, S. Lucey†, P. Lucey‡, S. Sridharan∗, and J.F. Cohn‡ +</td></tr><tr><td>75cd81d2513b7e41ac971be08bbb25c63c37029a</td><td></td></tr><tr><td>75e5ba7621935b57b2be7bf4a10cad66a9c445b9</td><td></td></tr><tr><td>75859ac30f5444f0d9acfeff618444ae280d661d</td><td>Multibiometric Cryptosystems based on Feature +<br/>Level Fusion +</td></tr><tr><td>758d7e1be64cc668c59ef33ba8882c8597406e53</td><td>IEEE TRANSACTIONS ON AFFECTIVE COMPUTING +<br/>AffectNet: A Database for Facial Expression, +<br/>Valence, and Arousal Computing in the Wild +</td></tr><tr><td>754f7f3e9a44506b814bf9dc06e44fecde599878</td><td>Quantized Densely Connected U-Nets for +<br/>Efficient Landmark Localization +</td></tr><tr><td>75249ebb85b74e8932496272f38af274fbcfd696</td><td>Face Identification in Large Galleries +<br/>Smart Surveillance Interest Group, Department of Computer Science +<br/>Universidade Federal de Minas Gerais, Belo Horizonte, Brazil +</td></tr><tr><td>81a142c751bf0b23315fb6717bc467aa4fdfbc92</td><td>978-1-5090-4117-6/17/$31.00 ©2017 IEEE +<br/>1767 +<br/>ICASSP 2017 +</td></tr><tr><td>8147ee02ec5ff3a585dddcd000974896cb2edc53</td><td>Angular Embedding: +<br/>A Robust Quadratic Criterion +<br/>Stella X. Yu, Member, +<br/>IEEE +</td></tr><tr><td>8199803f476c12c7f6c0124d55d156b5d91314b6</td><td>The iNaturalist Species Classification and Detection Dataset +<br/>1Caltech +<br/>2Google +<br/>3Cornell Tech +<br/>4iNaturalist +</td></tr><tr><td>81831ed8e5b304e9d28d2d8524d952b12b4cbf55</td><td></td></tr><tr><td>81b2a541d6c42679e946a5281b4b9dc603bc171c</td><td>Universit¨at Ulm | 89069 Ulm | Deutschland +<br/>Fakult¨at f¨ur Ingenieurwissenschaften und Informatik +<br/>Institut f¨ur Neuroinformatik +<br/>Direktor: Prof. Dr. G¨unther Palm +<br/>Semi-Supervised Learning with Committees: +<br/>Exploiting Unlabeled Data Using Ensemble +<br/>Learning Algorithms +<br/>Dissertation zur Erlangung des Doktorgrades +<br/>Doktor der Naturwissenschaften (Dr. rer. nat.) +<br/>der Fakult¨at f¨ur Ingenieurwissenschaften und Informatik +<br/>der Universit¨at Ulm +<br/>vorgelegt von +<br/>aus Kairo, ¨Agypten +<br/>Ulm, Deutschland +<br/>2010 +</td></tr><tr><td>8160b3b5f07deaa104769a2abb7017e9c031f1c1</td><td>683 +<br/>Exploiting Discriminant Information in Nonnegative +<br/>Matrix Factorization With Application +<br/>to Frontal Face Verification +</td></tr><tr><td>816eff5e92a6326a8ab50c4c50450a6d02047b5e</td><td>fLRR: Fast Low-Rank Representation Using +<br/>Frobenius Norm +<br/>Low Rank Representation (LRR) intends to find the representation +<br/>with lowest-rank of a given data set, which can be formulated as a +<br/>rank minimization problem. Since the rank operator is non-convex and +<br/>discontinuous, most of the recent works use the nuclear norm as a convex +<br/>relaxation. This letter theoretically shows that under some conditions, +<br/>Frobenius-norm-based optimization problem has an unique solution that +<br/>is also a solution of the original LRR optimization problem. In other +<br/>words, it is feasible to apply Frobenius-norm as a surrogate of the +<br/>nonconvex matrix rank function. This replacement will largely reduce the +<br/>time-costs for obtaining the lowest-rank solution. Experimental results +<br/>show that our method (i.e., fast Low Rank Representation, fLRR), +<br/>performs well in terms of accuracy and computation speed in image +<br/>clustering and motion segmentation compared with nuclear-norm-based +<br/>LRR algorithm. +<br/>Introduction: Given a data set X ∈ Rm×n(m < n) composed of column +<br/>vectors, let A be a data set composed of vectors with the same dimension +<br/>as those in X. Both X and A can be considered as matrices. A linear +<br/>representation of X with respect to A is a matrix Z that satisfies the +<br/>equation X = AZ. The data set A is called a dictionary. In general, this +<br/>linear matrix equation will have infinite solutions, and any solution can be +<br/>considered to be a representation of X associated with the dictionary A. To +<br/>obtain an unique Z and explore the latent structure of the given data set, +<br/>various assumptions could be enforced over Z. +<br/>Liu et al. recently proposed Low Rank Representation (LRR) [1] by +<br/>assuming that data are approximately sampled from an union of low-rank +<br/>subspaces. Mathematically, LRR aims at solving +<br/>min rank(Z) +<br/>s.t. X = AZ, +<br/>(1) +<br/>where rank(Z) could be defined as the number of nonzero eigenvalues of +<br/>the matrix Z. Clearly, (1) is non-convex and discontinuous, whose convex +<br/>relaxation is as follows, +<br/>min kZk∗ +<br/>s.t. X = AZ, +<br/>(2) +<br/>where kZk∗ is the nuclear norm, which is a convex and continuous +<br/>optimization problem. +<br/>Considering the possible corruptions, the objective function of LRR is +<br/>min kZk∗ + λkEkp +<br/>s.t. X = AZ + E, +<br/>(3) +<br/>where k · kp could be ℓ1-norm for describing sparse corruption or ℓ2,1- +<br/>norm for characterizing sample-specified corruption. +<br/>The above nuclear-norm-based optimization problems are generally +<br/>solved using Augmented Lagrange Multiplier algorithm (ALM) [2] which +<br/>requires repeatedly performing Single Value Decomposition (SVD) over +<br/>Z. Hence, this optimization program is inefficient. +<br/>Beyond the nuclear-norm, do other norms exist that can be used as +<br/>a surrogates for rank-minimization problem in LRR? Can we develop +<br/>a fast algorithm to calculate LRR? This letter addresses these problems +<br/>by theoretically showing the equivalence between the solutions of a +<br/>Frobenius-norm-based problem and the original LRR problem. And we +<br/>further develop fast Low Rank Representation (fLRR) based on the +<br/>theoretical results. +<br/>Theoretical Analysis: In the following analyses, Theorem 1 and +<br/>Theorem 3 prove that Frobenius-norm-based problem is a surrogate of +<br/>the rank-minimization problem of LRR in the case of clean data and +<br/>corrupted ones, respectively. Theorem 2 shows that our Frobenius-norm- +<br/>based method could produce a block-diagonal Z under some conditions. +<br/>This property is helpful to subspace clustering. +<br/>Let A ∈ Rm×n be a matrix with rank r. The full SVD and skinny +<br/>SVD of A are A = U ΣV T and A = UrΣrV T +<br/>r , where U and V are two +<br/>orthogonal matrices with the size of m × m and n × n, respectively. In +<br/>addition, Σ is an m × n rectangular diagonal matrix, its diagonal elements +<br/>are nonnegative real numbers. Σr is a r × r diagonal matrix with singular +<br/>values located on the diagonal in decreasing order, Ur and Vr consist of the +<br/>first r columns of U and V , respectively. Clearly, Ur and Vr are column +<br/>orthogonal matrices, i.e., U T +<br/>r Vr = Ir, where Ir denotes the +<br/>r Ur = Ir, V T +<br/>identity matrix with the size of r × r. The pseudoinverse of A is defined +<br/>by A† = VrΣ−1 +<br/>r U T +<br/>r . +<br/>Given a matrix M ∈ Rm×n, the Frobenius norm of M is defined by +<br/>kM kF =ptrace (M T M ) =qPmin{m,n} +<br/>value of M . Clearly, kM kF = 0 if and only if M = 0. +<br/>i=1 +<br/>σ2 +<br/>i , where σi is a singular +<br/>Lemma 1: Suppose P is a column orthogonal matrix, i.e., P T P = I. Then, +<br/>kP M kF = kM kF . +<br/>Lemma 2: For the matrices M and N with same number of columns, it +<br/>holds that +<br/>= kM k2 +<br/>F + kN k2 +<br/>F . +<br/>(4) +<br/>N (cid:21)(cid:13)(cid:13)(cid:13)(cid:13) +<br/>(cid:13)(cid:13)(cid:13)(cid:13) +<br/>(cid:20) M +<br/>The proofs of the above two lemmas are trivial. +<br/>Theorem 1: +<br/>minimization problem +<br/>Suppose +<br/>that X ∈ span{A}, +<br/>the Frobenius norm +<br/>min kZkF +<br/>s.t. X = AZ, +<br/>(5) +<br/>has an unique solution Z ∗ = A†X which is also the lowest-rank solution +<br/>of LRR in terms of (1). +<br/>Proof: Let the full and skinny SVDs of A be A = U ΣV T and A = +<br/>r U T +<br/>UrΣrV T +<br/>r . +<br/>r , respectively. Then, the pseudoinverse of A is A† = VrΣ−1 +<br/>Defining Vc by V T =(cid:20) V T +<br/>V T +<br/>(cid:21) and V T +<br/>c Vr = 0. Moreover, it can be easily +<br/>checked that Z ∗ satisfies X = AZ ∗ owing to X ∈ span{A}. +<br/>To prove that Z ∗ is the unique solution of the optimization problem +<br/>(5), two steps are required. First, we will prove that, for any solution Z of +<br/>X = AZ, it must hold that kZkF ≥ kZ ∗kF . Using Lemma 1, we have +<br/>kZkF = (cid:13)(cid:13)(cid:13)(cid:13) +<br/>= (cid:13)(cid:13)(cid:13)(cid:13) +<br/>V T +<br/>(cid:20) V T +<br/>(cid:20) V T +<br/>(cid:21) [Z ∗ + (Z − Z ∗)](cid:13)(cid:13)(cid:13)(cid:13)F +<br/>c (Z − Z ∗) (cid:21)(cid:13)(cid:13)(cid:13)(cid:13)F +<br/>r (Z − Z ∗) +<br/>r Z ∗ + V T +<br/>c Z ∗ + V T +<br/>V T +<br/>As A (Z − Z ∗) = 0, +<br/>r (Z − Z ∗) = 0. Denote B = Σ−1 +<br/>V T +<br/>V T +<br/>c Vr = 0, we have V T +<br/>i.e., UrΣrV T +<br/>r U T +<br/>c VrB = 0. Then, +<br/>r (Z − Z ∗) = 0, +<br/>r X, +<br/>follows that +<br/>then Z ∗ = VrB. Because +<br/>it +<br/>c Z ∗ = V T +<br/>(cid:20) +<br/>kZkF =(cid:13)(cid:13)(cid:13)(cid:13) +<br/>V T +<br/>c (Z − Z ∗) (cid:21)(cid:13)(cid:13)(cid:13)(cid:13)F +<br/>By Lemma 2, +<br/>kZk2 +<br/>F = kBk2 +<br/>F + kV T +<br/>c (Z − Z ∗)k2 +<br/>F , +<br/>then, kZkF ≥ kBkF . +<br/>By Lemma 1, +<br/>kBkF = kVrBkF = kZ ∗kF , +<br/>(6) +<br/>(7) +<br/>(8) +<br/>thus, kZkF ≥ kZ ∗kF for any solution Z of X = AZ. +<br/>In the second step, we will prove that if there exists another solution Z +<br/>of (5), Z = Z ∗ must hold. Clearly, Z is a solution of (5) which implies that +<br/>X = AZ and kZkF = kZ ∗kF . From (7) and (8), +<br/>kZk2 +<br/>F + kV T +<br/>F = kZ ∗k2 +<br/>Since kZkF = kZ ∗kF , +<br/>c (Z − Z ∗) k2 +<br/>F . +<br/>c (Z − Z ∗) kF = 0, +<br/>r (Z − Z ∗) = 0, this gives +<br/>and so V T +<br/>V T (Z − Z ∗) = 0. Because V is an orthogonal matrix, it must hold +<br/>that Z = Z ∗. The above proves that Z ∗ is the unique solution of the +<br/>optimization problem (5). +<br/>c (Z − Z ∗) = 0. Together with V T +<br/>it must hold that kV T +<br/>(9) +<br/>Next, we prove that Z ∗ is also a solution of the LRR optimization +<br/>problem (1). Clearly, for any solution Z of X = AZ, +<br/>it holds that +<br/>rank(Z) ≥ rank(AZ) = rank(X). On the other hand, rank(Z ∗) = +<br/>rank(A†X) ≤ rank(X). Thus, rank(Z ∗) = rank(X). This shows that +<br/>Z ∗ is the lowest-rank solution of the LRR optimization problem (1). The +<br/>proof is complete. +<br/>(cid:4) +<br/>In the following, Theorem 2 will show that the optimal Z of (5) will +<br/>be block-diagonal if the data are sampled from a set of independent +<br/>subspaces {S1, S2, · · · , Sk}, where the dimensionality of Si is ri and +<br/>i = {1, 2, · · · , k}. Note that, {S1, S2, · · · , Sk} are independent if and +<br/>only if SiTPj6=i Sj = {0}. Suppose that X = [X1, X2, · · · , Xk] and +<br/>A = [A1, A2, · · · , Ak], where Ai and Xi contain mi and ni data points +<br/>ELECTRONICS LETTERS 12th December 2011 Vol. 00 No. 00 +</td></tr><tr><td>8149c30a86e1a7db4b11965fe209fe0b75446a8c</td><td>Semi-Supervised Multiple Instance Learning based +<br/>Domain Adaptation for Object Detection +<br/>Siemens Corporate Research +<br/>Siemens Corporate Research +<br/>Siemens Corporate Research +<br/>Amit Kale +<br/>Bangalore +<br/>Bangalore +<br/>{chhaya.methani, +<br/>Bangalore +<br/>rahul.thota, +</td></tr><tr><td>86b69b3718b9350c9d2008880ce88cd035828432</td><td>Improving Face Image Extraction by Using Deep Learning Technique +<br/>National Library of Medicine, NIH, Bethesda, MD +</td></tr><tr><td>86904aee566716d9bef508aa9f0255dc18be3960</td><td>Learning Anonymized Representations with +<br/>Adversarial Neural Networks +</td></tr><tr><td>867e709a298024a3c9777145e037e239385c0129</td><td> INTERNATIONAL JOURNAL +<br/> OF PROFESSIONAL ENGINEERING STUDIES Volume VIII /Issue 2 / FEB 2017 +<br/>ANALYTICAL REPRESENTATION OF UNDERSAMPLED FACE +<br/>RECOGNITION APPROACH BASED ON DICTIONARY LEARNING +<br/>AND SPARSE REPRESENTATION +<br/>(M.Tech)1, Assistant Professor2, Assistant Professor3, HOD of CSE Department4 +</td></tr><tr><td>86c053c162c08bc3fe093cc10398b9e64367a100</td><td>Cascade of Forests for Face Alignment +</td></tr><tr><td>86b985b285c0982046650e8d9cf09565a939e4f9</td><td></td></tr><tr><td>861802ac19653a7831b314cd751fd8e89494ab12</td><td>Time-of-Flight and Depth Imaging. Sensors, Algorithms +<br/>and Applications: Dagstuhl Seminar 2012 and GCPR +<br/>Workshop on Imaging New Modalities (Lecture ... Vision, +<br/>Pattern Recognition, and Graphics) +<br/>Publisher: Springer; 2013 edition +<br/>(November 8, 2013) +<br/>Language: English +<br/>Pages: 320 +<br/>ISBN: 978-3642449635 +<br/>Size: 20.46 MB +<br/>Format: PDF / ePub / Kindle +<br/>Cameras for 3D depth imaging, using +<br/>either time-of-flight (ToF) or +<br/>structured light sensors, have received +<br/>a lot of attention recently and have +<br/>been improved considerably over the +<br/>last few years. The present +<br/>techniques... +</td></tr><tr><td>861b12f405c464b3ffa2af7408bff0698c6c9bf0</td><td>International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169 +<br/>Volume: 3 Issue: 5 +<br/> 3337 - 3342 +<br/>_______________________________________________________________________________________________ +<br/>An Effective Technique for Removal of Facial Dupilcation by SBFA +<br/>Computer Department, +<br/>GHRCEM, +<br/>Pune, India +<br/>Computer Department, +<br/>GHRCEM, +<br/> Pune, India +</td></tr><tr><td>86e1bdbfd13b9ed137e4c4b8b459a3980eb257f6</td><td>The Kinetics Human Action Video Dataset +<br/>Jo˜ao Carreira +<br/>Paul Natsev +</td></tr><tr><td>86b105c3619a433b6f9632adcf9b253ff98aee87</td><td>1424403677/06/$20.00 ©2006 IEEE +<br/>1013 +<br/>ICME 2006 +</td></tr><tr><td>86b51bd0c80eecd6acce9fc538f284b2ded5bcdd</td><td></td></tr><tr><td>8699268ee81a7472a0807c1d3b1db0d0ab05f40d</td><td></td></tr><tr><td>869583b700ecf33a9987447aee9444abfe23f343</td><td></td></tr><tr><td>72a00953f3f60a792de019a948174bf680cd6c9f</td><td>Stat Comput (2007) 17:57–70 +<br/>DOI 10.1007/s11222-006-9004-9 +<br/>Understanding the role of facial asymmetry in human face +<br/>identification +<br/>Received: May 2005 / Accepted: September 2006 / Published online: 30 January 2007 +<br/>C(cid:1) Springer Science + Business Media, LLC 2007 +</td></tr><tr><td>726b8aba2095eef076922351e9d3a724bb71cb51</td><td></td></tr><tr><td>721b109970bf5f1862767a1bec3f9a79e815f79a</td><td></td></tr><tr><td>72ecaff8b57023f9fbf8b5b2588f3c7019010ca7</td><td>Facial Keypoints Detection +</td></tr><tr><td>72591a75469321074b072daff80477d8911c3af3</td><td>Group Component Analysis for Multi-block Data: +<br/>Common and Individual Feature Extraction +</td></tr><tr><td>729dbe38538fbf2664bc79847601f00593474b05</td><td></td></tr><tr><td>729a9d35bc291cc7117b924219bef89a864ce62c</td><td>Recognizing Material Properties from Images +</td></tr><tr><td>721d9c387ed382988fce6fa864446fed5fb23173</td><td></td></tr><tr><td>72c0c8deb9ea6f59fde4f5043bff67366b86bd66</td><td>Age progression in Human Faces : A Survey +</td></tr><tr><td>445461a34adc4bcdccac2e3c374f5921c93750f8</td><td>Emotional Expression Classification using Time-Series Kernels∗ +</td></tr><tr><td>4414a328466db1e8ab9651bf4e0f9f1fe1a163e4</td><td>1164 +<br/>© EURASIP, 2010 ISSN 2076-1465 +<br/>18th European Signal Processing Conference (EUSIPCO-2010) +<br/>INTRODUCTION +</td></tr><tr><td>442f09ddb5bb7ba4e824c0795e37cad754967208</td><td></td></tr><tr><td>446a99fdedd5bb32d4970842b3ce0fc4f5e5fa03</td><td>A Pose-Adaptive Constrained Local Model For +<br/>Accurate Head Pose Tracking +<br/>Eikeo +<br/>11 rue Leon Jouhaux, +<br/>F-75010, Paris, France +<br/>Sorbonne Universit´es +<br/>UPMC Univ Paris 06 +<br/>CNRS UMR 7222, ISIR +<br/>F-75005, Paris, France +<br/>Eikeo +<br/>11 rue Leon Jouhaux, +<br/>F-75010, Paris, France +</td></tr><tr><td>44b1399e8569a29eed0d22d88767b1891dbcf987</td><td>This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. +<br/>IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE +<br/>Learning Multi-modal Latent Attributes +</td></tr><tr><td>446dc1413e1cfaee0030dc74a3cee49a47386355</td><td>Recent Advances in Zero-shot Recognition +</td></tr><tr><td>44a3ec27f92c344a15deb8e5dc3a5b3797505c06</td><td>A Taxonomy of Part and Attribute Discovery +<br/>Techniques +</td></tr><tr><td>44aeda8493ad0d44ca1304756cc0126a2720f07b</td><td>Face Alive Icons +</td></tr><tr><td>449b1b91029e84dab14b80852e35387a9275870e</td><td></td></tr><tr><td>44078d0daed8b13114cffb15b368acc467f96351</td><td></td></tr><tr><td>44dd150b9020b2253107b4a4af3644f0a51718a3</td><td>An Analysis of the Sensitivity of Active Shape +<br/>Models to Initialization when Applied to Automatic +<br/>Facial Landmarking +</td></tr><tr><td>447d8893a4bdc29fa1214e53499ffe67b28a6db5</td><td></td></tr><tr><td>44f65e3304bdde4be04823fd7ca770c1c05c2cef</td><td>SIViP +<br/>DOI 10.1007/s11760-009-0125-4 +<br/>ORIGINAL PAPER +<br/>On the use of phase of the Fourier transform for face recognition +<br/>under variations in illumination +<br/>Received: 17 November 2008 / Revised: 20 February 2009 / Accepted: 7 July 2009 +<br/>© Springer-Verlag London Limited 2009 +</td></tr><tr><td>44eb4d128b60485377e74ffb5facc0bf4ddeb022</td><td></td></tr><tr><td>448ed201f6fceaa6533d88b0b29da3f36235e131</td><td></td></tr><tr><td>447a5e1caf847952d2bb526ab2fb75898466d1bc</td><td>Under review as a conference paper at ICLR 2018 +<br/>LEARNING NON-LINEAR TRANSFORM WITH DISCRIM- +<br/>INATIVE AND MINIMUM INFORMATION LOSS PRIORS +<br/>Anonymous authors +<br/>Paper under double-blind review +</td></tr><tr><td>2a7bca56e2539c8cf1ae4e9da521879b7951872d</td><td>Exploiting Unrelated Tasks in Multi-Task Learning +<br/>Anonymous Author 1 +<br/>Unknown Institution 1 +<br/>Anonymous Author 2 +<br/>Unknown Institution 2 +<br/>Anonymous Author 3 +<br/>Unknown Institution 3 +</td></tr><tr><td>2aaa6969c03f435b3ea8431574a91a0843bd320b</td><td></td></tr><tr><td>2ad7cef781f98fd66101fa4a78e012369d064830</td><td></td></tr><tr><td>2ad29b2921aba7738c51d9025b342a0ec770c6ea</td><td></td></tr><tr><td>2a6bba2e81d5fb3c0fd0e6b757cf50ba7bf8e924</td><td></td></tr><tr><td>2aec012bb6dcaacd9d7a1e45bc5204fac7b63b3c</td><td>Robust Registration and Geometry Estimation from Unstructured +<br/>Facial Scans +</td></tr><tr><td>2ae139b247057c02cda352f6661f46f7feb38e45</td><td>Combining Modality Specific Deep Neural Networks for +<br/>Emotion Recognition in Video +<br/>1École Polytechique de Montréal, Université de Montréal, Montréal, Canada +<br/>2Laboratoire d’Informatique des Systèmes Adaptatifs, Université de Montréal, Montréal, Canada +</td></tr><tr><td>2a5903bdb3fdfb4d51f70b77f16852df3b8e5f83</td><td>121 +<br/>The Effect of Computer-Generated Descriptions +<br/>on Photo-Sharing Experiences of People With +<br/>Visual Impairments +<br/>Like sighted people, visually impaired people want to share photographs on social networking services, but +<br/>find it difficult to identify and select photos from their albums. We aimed to address this problem by +<br/>incorporating state-of-the-art computer-generated descriptions into Facebook’s photo-sharing feature. We +<br/>interviewed 12 visually impaired participants to understand their photo-sharing experiences and designed a +<br/>photo description feature for the Facebook mobile application. We evaluated this feature with six +<br/>participants in a seven-day diary study. We found that participants used the descriptions to recall and +<br/>organize their photos, but they hesitated to upload photos without a sighted person’s input. In addition to +<br/>basic information about photo content, participants wanted to know more details about salient objects and +<br/>people, and whether the photos reflected their personal aesthetic. We discuss these findings from the lens of +<br/>self-disclosure and self-presentation theories and propose new computer vision research directions that will +<br/>better support visual content sharing by visually impaired people. +<br/>CCS Concepts: • Information interfaces and presentations → Multimedia and information systems; • +<br/>Social and professional topics → People with disabilities +<br/>KEYWORDS +<br/>Visual impairments; computer-generated descriptions; SNSs; photo sharing; self-disclosure; self-presentation +<br/>ACM Reference format: +<br/>The Effect of Computer-Generated Descriptions On Photo-Sharing Experiences of People With Visual +<br/>Impairments. Proc. ACM Hum.-Comput. Interact. 1, CSCW. 121 (November 2017), 22 pages. +<br/>DOI: 10.1145/3134756 +<br/>1 INTRODUCTION +<br/>Sharing memories and experiences via photos is a common way to engage with others on social networking +<br/>services (SNSs) [39,46,51]. For instance, Facebook users uploaded more than 350 million photos a day [24] +<br/>and Twitter, which initially supported only text in tweets, now has more than 28.4% of tweets containing +<br/>images [39]. Visually impaired people (both blind and low vision) have a strong presence on SNS and are +<br/>interested in sharing photos [50]. They take photos for the same reasons that sighted people do: sharing +<br/>daily moments with their sighted friends and family [30,32]. A prior study showed that visually impaired +<br/>people shared a relatively large number of photos on Facebook—only slightly less than their sighted +<br/>counterparts [50]. +<br/> +<br/> PACM on Human-Computer Interaction, Vol. 1, No. 2, Article 121. Publication date: November 2017 +</td></tr><tr><td>2a02355c1155f2d2e0cf7a8e197e0d0075437b19</td><td></td></tr><tr><td>2aea27352406a2066ddae5fad6f3f13afdc90be9</td><td></td></tr><tr><td>2ad0ee93d029e790ebb50574f403a09854b65b7e</td><td>Acquiring Linear Subspaces for Face +<br/>Recognition under Variable Lighting +<br/>David Kriegman, Senior Member, IEEE +</td></tr><tr><td>2ff9618ea521df3c916abc88e7c85220d9f0ff06</td><td>Facial Tic Detection Using Computer Vision +<br/>Christopher D. Leveille +<br/>March 20, 2014 +</td></tr><tr><td>2fda461869f84a9298a0e93ef280f79b9fb76f94</td><td>OpenFace: an open source facial behavior analysis toolkit +<br/>Tadas Baltruˇsaitis +</td></tr><tr><td>2fdce3228d384456ea9faff108b9c6d0cf39e7c7</td><td></td></tr><tr><td>2f7e9b45255c9029d2ae97bbb004d6072e70fa79</td><td>Noname manuscript No. +<br/>(will be inserted by the editor) +<br/>cvpaper.challenge in 2015 +<br/>A review of CVPR2015 and DeepSurvey +<br/>Nakamura +<br/>Received: date / Accepted: date +</td></tr><tr><td>2f489bd9bfb61a7d7165a2f05c03377a00072477</td><td>JIA, YANG: STRUCTURED SEMI-SUPERVISED FOREST +<br/>Structured Semi-supervised Forest for +<br/>Facial Landmarks Localization with Face +<br/>Mask Reasoning +<br/>1 Department of Computer Science +<br/>The Univ. of Hong Kong, HK +<br/>2 School of EECS +<br/>Queen Mary Univ. of London, UK +<br/>Angran Lin1 +</td></tr><tr><td>2f16459e2e24dc91b3b4cac7c6294387d4a0eacf</td><td></td></tr><tr><td>2f59f28a1ca3130d413e8e8b59fb30d50ac020e2</td><td>Children Gender Recognition Under Unconstrained +<br/>Conditions Based on Contextual Information +<br/>Joint Research Centre, European Commission, Ispra, Italy +</td></tr><tr><td>2f88d3189723669f957d83ad542ac5c2341c37a5</td><td>Downloaded From: https://www.spiedigitallibrary.org/journals/Journal-of-Electronic-Imaging on 9/13/2018 +<br/>Terms of Use: https://www.spiedigitallibrary.org/terms-of-use +<br/>Attribute-correlatedlocalregionsfordeeprelativeattributeslearningFenZhangXiangweiKongZeJiaFenZhang,XiangweiKong,ZeJia,“Attribute-correlatedlocalregionsfordeeprelativeattributeslearning,”J.Electron.Imaging27(4),043021(2018),doi:10.1117/1.JEI.27.4.043021. </td></tr><tr><td>2fda164863a06a92d3a910b96eef927269aeb730</td><td>Names and Faces in the News +<br/>Computer Science Division +<br/>U.C. Berkeley +<br/>Berkeley, CA 94720 +</td></tr><tr><td>2fa057a20a2b4a4f344988fee0a49fce85b0dc33</td><td></td></tr><tr><td>2f8ef26bfecaaa102a55b752860dbb92f1a11dc6</td><td>A Graph Based Approach to Speaker Retrieval in Talk +<br/>Show Videos with Transcript-Based Supervision +</td></tr><tr><td>2f184c6e2c31d23ef083c881de36b9b9b6997ce9</td><td>Polichotomies on Imbalanced Domains +<br/>by One-per-Class Compensated Reconstruction Rule +<br/>Integrated Research Centre, Universit´a Campus Bio-Medico of Rome, Rome, Italy +</td></tr><tr><td>2f9c173ccd8c1e6b88d7fb95d6679838bc9ca51d</td><td></td></tr><tr><td>2f8183b549ec51b67f7dad717f0db6bf342c9d02</td><td></td></tr><tr><td>2fa1fc116731b2b5bb97f06d2ac494cb2b2fe475</td><td>A novel approach to personal photo album representation +<br/>and management +<br/>Universit`a di Palermo - Dipartimento di Ingegneria Informatica +<br/>Viale delle Scienze, 90128, Palermo, Italy +</td></tr><tr><td>2f882ceaaf110046e63123b495212d7d4e99f33d</td><td>High Frequency Component Compensation based Super-resolution +<br/>Algorithm for Face Video Enhancement +<br/>CVRR Lab, UC San Diego, La Jolla, CA 92093, USA +</td></tr><tr><td>2f95340b01cfa48b867f336185e89acfedfa4d92</td><td>Face Expression Recognition with a 2-Channel +<br/>Convolutional Neural Network +<br/><b></b><br/>Vogt-K¨olln-Straße 30, 22527 Hamburg, Germany +<br/>http://www.informatik.uni-hamburg.de/WTM/ +</td></tr><tr><td>2fea258320c50f36408032c05c54ba455d575809</td><td></td></tr><tr><td>2faa09413162b0a7629db93fbb27eda5aeac54ca</td><td>NISTIR 7674 +<br/>Quantifying How Lighting and Focus +<br/>Affect Face Recognition Performance +<br/>Phillips, P. J. +<br/>Beveridge, J. R. +<br/>Draper, B. +<br/>Bolme, D. +<br/>Givens, G. H. +<br/>Lui, Y. M. +<br/>1 +</td></tr><tr><td>433bb1eaa3751519c2e5f17f47f8532322abbe6d</td><td></td></tr><tr><td>4300fa1221beb9dc81a496cd2f645c990a7ede53</td><td></td></tr><tr><td>439ac8edfa1e7cbc65474cab544a5b8c4c65d5db</td><td>SIViP (2011) 5:401–413 +<br/>DOI 10.1007/s11760-011-0244-6 +<br/>ORIGINAL PAPER +<br/>Face authentication with undercontrolled pose and illumination +<br/>Received: 15 September 2010 / Revised: 14 December 2010 / Accepted: 17 February 2011 / Published online: 7 August 2011 +<br/>© Springer-Verlag London Limited 2011 +</td></tr><tr><td>43f6953804964037ff91a4f45d5b5d2f8edfe4d5</td><td>Multi-Feature Fusion in Advanced Robotics Applications +<br/>Institut für Informatik +<br/>Technische Universität München +<br/>D-85748 Garching, Germany +</td></tr><tr><td>439ec47725ae4a3660e509d32828599a495559bf</td><td>Facial Expressions Tracking and Recognition: Database Protocols for Systems Validation +<br/>and Evaluation +</td></tr><tr><td>43a03cbe8b704f31046a5aba05153eb3d6de4142</td><td>Towards Robust Face Recognition from Video +<br/>Image Science and Machine Vision Group +<br/>Oak Ridge National Laboratory +<br/>Oak Ridge, TN 37831-6010 +</td></tr><tr><td>43836d69f00275ba2f3d135f0ca9cf88d1209a87</td><td>Ozaki et al. IPSJ Transactions on Computer Vision and +<br/>Applications (2017) 9:20 +<br/>DOI 10.1186/s41074-017-0030-7 +<br/>IPSJ Transactions on Computer +<br/>Vision and Applications +<br/>RESEARCH PAPER +<br/>Open Access +<br/>Effective hyperparameter optimization +<br/>using Nelder-Mead method in deep learning +</td></tr><tr><td>43aa40eaa59244c233f83d81f86e12eba8d74b59</td><td></td></tr><tr><td>4362368dae29cc66a47114d5ffeaf0534bf0159c</td><td>UACEE International Journal of Artificial Intelligence and Neural Networks ISSN:- 2250-3749 (online) +<br/>Performance Analysis of FDA Based Face +<br/>Recognition Using Correlation, ANN and SVM +<br/>Department of Computer Engineering +<br/>Department of Computer Engineering +<br/>Department of Computer Engineering +<br/>Anand, INDIA +<br/>Anand, INDIA +<br/>Anand, INDIA +</td></tr><tr><td>43e268c118ac25f1f0e984b57bc54f0119ded520</td><td></td></tr><tr><td>43476cbf2a109f8381b398e7a1ddd794b29a9a16</td><td>A Practical Transfer Learning Algorithm for Face Verification +<br/>David Wipf +</td></tr><tr><td>4353d0dcaf450743e9eddd2aeedee4d01a1be78b</td><td>Learning Discriminative LBP-Histogram Bins +<br/>for Facial Expression Recognition +<br/>Philips Research, High Tech Campus 36, Eindhoven 5656 AE, The Netherlands +</td></tr><tr><td>437a720c6f6fc1959ba95e48e487eb3767b4e508</td><td></td></tr><tr><td>436d80cc1b52365ed7b2477c0b385b6fbbb51d3b</td><td></td></tr><tr><td>43b8b5eeb4869372ef896ca2d1e6010552cdc4d4</td><td>Large-scale Supervised Hierarchical Feature Learning for Face Recognition +<br/>Intel Labs China +</td></tr><tr><td>43ae4867d058453e9abce760ff0f9427789bab3a</td><td>951 +<br/>Graph Embedded Nonparametric Mutual +<br/>Information For Supervised +<br/>Dimensionality Reduction +</td></tr><tr><td>430c4d7ad76e51d83bbd7ec9d3f856043f054915</td><td></td></tr><tr><td>438b88fe40a6f9b5dcf08e64e27b2719940995e0</td><td>Building a Classi(cid:2)cation Cascade for Visual Identi(cid:2)cation from One Example +<br/>Computer Science, U.C. Berkeley +<br/>Computer Science, UMass Amherst +<br/>Computer Science, U.C. Berkeley +</td></tr><tr><td>43fb9efa79178cb6f481387b7c6e9b0ca3761da8</td><td>Mixture of Parts Revisited: Expressive Part Interactions for Pose Estimation +<br/>Anoop R Katti +<br/>IIT Madras +<br/>Chennai, India +<br/>IIT Madras +<br/>Chennai, India +</td></tr><tr><td>43d7d0d0d0e2d6cf5355e60c4fe5b715f0a1101a</td><td>Pobrane z czasopisma Annales AI- Informatica http://ai.annales.umcs.pl +<br/>Data: 04/05/2018 16:53:32 +<br/>U M CS +</td></tr><tr><td>889bc64c7da8e2a85ae6af320ae10e05c4cd6ce7</td><td>174 +<br/>Using Support Vector Machines to Enhance the +<br/>Performance of Bayesian Face Recognition +</td></tr><tr><td>8812aef6bdac056b00525f0642702ecf8d57790b</td><td>A Unified Features Approach to Human Face Image +<br/>Analysis and Interpretation +<br/>Department of Informatics, +<br/>Technische Universit¨at M¨unchen +<br/>85748 Garching, Germany +</td></tr><tr><td>881066ec43bcf7476479a4146568414e419da804</td><td>From Traditional to Modern : Domain Adaptation for +<br/>Action Classification in Short Social Video Clips +<br/>Center for Visual Information Technology, IIIT Hyderabad, India +</td></tr><tr><td>8813368c6c14552539137aba2b6f8c55f561b75f</td><td>Trunk-Branch Ensemble Convolutional Neural +<br/>Networks for Video-based Face Recognition +</td></tr><tr><td>883006c0f76cf348a5f8339bfcb649a3e46e2690</td><td>Weakly Supervised Pain Localization using Multiple Instance Learning +</td></tr><tr><td>88f2952535df5859c8f60026f08b71976f8e19ec</td><td>A neural network framework for face +<br/>recognition by elastic bunch graph matching +</td></tr><tr><td>8818b12aa0ff3bf0b20f9caa250395cbea0e8769</td><td>Fashion Conversation Data on Instagram +<br/>∗Graduate School of Culture Technology, KAIST, South Korea +<br/>†Department of Communication Studies, UCLA, USA +</td></tr><tr><td>8878871ec2763f912102eeaff4b5a2febfc22fbe</td><td>3781 +<br/>Human Action Recognition in Unconstrained +<br/>Videos by Explicit Motion Modeling +</td></tr><tr><td>8855d6161d7e5b35f6c59e15b94db9fa5bbf2912</td><td>COGNITION IN PREGNANCY AND THE POSTPARTUM PERIOD +</td></tr><tr><td>88bee9733e96958444dc9e6bef191baba4fa6efa</td><td>Extending Face Identification to +<br/>Open-Set Face Recognition +<br/>Department of Computer Science +<br/>Universidade Federal de Minas Gerais +<br/>Belo Horizonte, Brazil +</td></tr><tr><td>88fd4d1d0f4014f2b2e343c83d8c7e46d198cc79</td><td>978-1-4799-9988-0/16/$31.00 ©2016 IEEE +<br/>2697 +<br/>ICASSP 2016 +</td></tr><tr><td>9fa1be81d31fba07a1bde0275b9d35c528f4d0b8</td><td>Identifying Persons by Pictorial and +<br/>Contextual Cues +<br/>Nicholas Leonard Pi¨el +<br/>Thesis submitted for the degree of Master of Science +<br/>Supervisor: +<br/>April 2009 +</td></tr><tr><td>9f094341bea610a10346f072bf865cb550a1f1c1</td><td>Recognition and Volume Estimation of Food Intake using a Mobile Device +<br/>Sarnoff Corporation +<br/>201 Washington Rd, +<br/>Princeton, NJ, 08540 +</td></tr><tr><td>6b333b2c6311e36c2bde920ab5813f8cfcf2b67b</td><td></td></tr><tr><td>6b9aa288ce7740ec5ce9826c66d059ddcfd8dba9</td><td></td></tr><tr><td>6b089627a4ea24bff193611e68390d1a4c3b3644</td><td>CROSS-POLLINATION OF NORMALISATION +<br/>TECHNIQUES FROM SPEAKER TO FACE +<br/>AUTHENTICATION USING GAUSSIAN +<br/>MIXTURE MODELS +<br/>Idiap-RR-03-2012 +<br/>JANUARY 2012 +<br/>Centre du Parc, Rue Marconi 19, P.O. Box 592, CH - 1920 Martigny +</td></tr><tr><td>6be0ab66c31023762e26d309a4a9d0096f72a7f0</td><td>Enhance Visual Recognition under Adverse +<br/>Conditions via Deep Networks +</td></tr><tr><td>6b18628cc8829c3bf851ea3ee3bcff8543391819</td><td>Face recognition based on subset selection via metric learning on manifold. +<br/>1058. [doi:10.1631/FITEE.1500085] +<br/>Face recognition based on subset +<br/>selection via metric learning on manifold +<br/>Key words: Face recognition, Sparse representation, Manifold structure, +<br/>Metric learning, Subset selection +<br/> ORCID: http://orcid.org/0000-0001-7441-4749 +<br/>Front Inform Technol & Electron Eng </td></tr><tr><td>6b1b43d58faed7b457b1d4e8c16f5f7e7d819239</td><td></td></tr><tr><td>6b35b15ceba2f26cf949f23347ec95bbbf7bed64</td><td></td></tr><tr><td>6b6493551017819a3d1f12bbf922a8a8c8cc2a03</td><td>Pose Normalization for Local Appearance-Based +<br/>Face Recognition +<br/>Computer Science Department, Universit¨at Karlsruhe (TH) +<br/>Am Fasanengarten 5, Karlsruhe 76131, Germany +<br/>http://isl.ira.uka.de/cvhci +</td></tr><tr><td>6bb630dfa797168e6627d972560c3d438f71ea99</td><td></td></tr><tr><td>0728f788107122d76dfafa4fb0c45c20dcf523ca</td><td>The Best of Both Worlds: Combining Data-independent and Data-driven +<br/>Approaches for Action Recognition +</td></tr><tr><td>071099a4c3eed464388c8d1bff7b0538c7322422</td><td>FACIAL EXPRESSION RECOGNITION IN THE WILD USING RICH DEEP FEATURES +<br/>Microsoft Advanced Technology labs, Microsoft Technology and Research, Cairo, Egypt +<br/> +</td></tr><tr><td>071af21377cc76d5c05100a745fb13cb2e40500f</td><td></td></tr><tr><td>0754e769eb613fd3968b6e267a301728f52358be</td><td>Towards a Watson That Sees: Language-Guided Action Recognition for +<br/>Robots +</td></tr><tr><td>0717b47ab84b848de37dbefd81cf8bf512b544ac</td><td>International Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 +<br/>International Conference on Humming Bird ( 01st March 2014) +<br/>RESEARCH ARTICLE +<br/> OPEN ACCESS +<br/>Robust Face Recognition and Tagging in Visual Surveillance +<br/>System +</td></tr><tr><td>0750a816858b601c0dbf4cfb68066ae7e788f05d</td><td>CosFace: Large Margin Cosine Loss for Deep Face Recognition +<br/>Tencent AI Lab +</td></tr><tr><td>0716e1ad868f5f446b1c367721418ffadfcf0519</td><td>Interactively Guiding Semi-Supervised +<br/>Clustering via Attribute-Based Explanations +<br/>Virginia Tech, Blacksburg, VA, USA +</td></tr><tr><td>073eaa49ccde15b62425cda1d9feab0fea03a842</td><td></td></tr><tr><td>0726a45eb129eed88915aa5a86df2af16a09bcc1</td><td>Introspective Perception: Learning to Predict Failures in Vision Systems +</td></tr><tr><td>38d56ddcea01ce99902dd75ad162213cbe4eaab7</td><td>Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence (IJCAI-17) +<br/>2648 +</td></tr><tr><td>389334e9a0d84bc54bcd5b94b4ce4c5d9d6a2f26</td><td>FACIAL PARAMETER EXTRACTION SYSTEM BASED ON ACTIVE CONTOURS +<br/>Universitat Politècnica de Catalunya, Barcelona, Spain +</td></tr><tr><td>380dd0ddd5d69adc52defc095570d1c22952f5cc</td><td></td></tr><tr><td>38679355d4cfea3a791005f211aa16e76b2eaa8d</td><td>Title +<br/>Evolutionary cross-domain discriminative Hessian Eigenmaps +<br/>Author(s) +<br/>Si, S; Tao, D; Chan, KP +<br/>Citation +<br/>1086 +<br/>Issued Date +<br/>2010 +<br/>URL +<br/>http://hdl.handle.net/10722/127357 +<br/>Rights +<br/>This work is licensed under a Creative Commons Attribution- +<br/>NonCommercial-NoDerivatives 4.0 International License.; ©2010 +<br/>IEEE. Personal use of this material is permitted. However, +<br/>permission to reprint/republish this material for advertising or +<br/>promotional purposes or for creating new collective works for +<br/>resale or redistribution to servers or lists, or to reuse any +<br/>copyrighted component of this work in other works must be +<br/>obtained from the IEEE. +</td></tr><tr><td>38682c7b19831e5d4f58e9bce9716f9c2c29c4e7</td><td>International Journal of Computer Trends and Technology (IJCTT) – Volume 18 Number 5 – Dec 2014 +<br/>Movie Character Identification Using Graph Matching +<br/>Algorithm +<br/>M.Tech Scholar, Dept of CSE, QISCET, ONGOLE, Dist: Prakasam, AP, India. +<br/>Associate Professor, Department of CSE, QISCET, ONGOLE, Dist: Prakasam, AP, India +</td></tr><tr><td>3803b91e784922a2dacd6a18f61b3100629df932</td><td>Temporal Multimodal Fusion +<br/>for Video Emotion Classification in the Wild +<br/>Orange Labs +<br/>Cesson-Sévigné, France +<br/>Orange Labs +<br/>Cesson-Sévigné, France +<br/>Normandie Univ., UNICAEN, +<br/>ENSICAEN, CNRS +<br/>Caen, France +</td></tr><tr><td>38eea307445a39ee7902c1ecf8cea7e3dcb7c0e7</td><td>Noname manuscript No. +<br/>(will be inserted by the editor) +<br/>Multi-distance Support Matrix Machine +<br/>Received: date / Accepted: date +</td></tr><tr><td>385750bcf95036c808d63db0e0b14768463ff4c6</td><td></td></tr><tr><td>384f972c81c52fe36849600728865ea50a0c4670</td><td>1 +<br/>Multi-Fold Gabor, PCA and ICA Filter +<br/>Convolution Descriptor for Face Recognition +<br/> +</td></tr><tr><td>380d5138cadccc9b5b91c707ba0a9220b0f39271</td><td>Deep Imbalanced Learning for Face Recognition +<br/>and Attribute Prediction +</td></tr><tr><td>38861d0d3a0292c1f54153b303b0d791cbba1d50</td><td></td></tr><tr><td>38192a0f9261d9727b119e294a65f2e25f72d7e6</td><td></td></tr><tr><td>00fb2836068042c19b5197d0999e8e93b920eb9c</td><td></td></tr><tr><td>0077cd8f97cafd2b389783858a6e4ab7887b0b6b</td><td>MAI et al.: ON THE RECONSTRUCTION OF DEEP FACE TEMPLATES +<br/>On the Reconstruction of Deep Face Templates +</td></tr><tr><td>00214fe1319113e6649435cae386019235474789</td><td>Bachelorarbeit im Fach Informatik +<br/>Face Recognition using +<br/>Distortion Models +<br/>Mathematik, Informatik und Naturwissenschaften der +<br/>RHEINISCH-WESTFÄLISCHEN TECHNISCHEN HOCHSCHULE AACHEN +<br/>Der Fakultät für +<br/>Lehrstuhl für Informatik VI +<br/>Prof. Dr.-Ing. H. Ney +<br/>vorgelegt von: +<br/>Matrikelnummer 252400 +<br/>Gutachter: +<br/>Prof. Dr.-Ing. H. Ney +<br/>Prof. Dr. B. Leibe +<br/>Betreuer: +<br/>September 2009 +</td></tr><tr><td>0004f72a00096fa410b179ad12aa3a0d10fc853c</td><td></td></tr><tr><td>00f0ed04defec19b4843b5b16557d8d0ccc5bb42</td><td></td></tr><tr><td>0037bff7be6d463785d4e5b2671da664cd7ef746</td><td>Author manuscript, published in "European Conference on Computer Vision (ECCV '10) 6311 (2010) 634--647" +<br/> DOI : 10.1007/978-3-642-15549-9_46 +</td></tr><tr><td>00d9d88bb1bdca35663946a76d807fff3dc1c15f</td><td>Subjects and Their Objects: Localizing Interactees for a +<br/>Person-Centric View of Importance +</td></tr><tr><td>00a967cb2d18e1394226ad37930524a31351f6cf</td><td>Fully-adaptive Feature Sharing in Multi-Task Networks with Applications in +<br/>Person Attribute Classification +<br/>UC San Diego +<br/>IBM Research +<br/>IBM Research +<br/>Binghamton Univeristy, SUNY +<br/>UC San Diego +<br/>Rogerio Feris +<br/>IBM Research +</td></tr><tr><td>00a3cfe3ce35a7ffb8214f6db15366f4e79761e3</td><td>Kinect for real-time emotion recognition via facial expressions. Frontiers of +<br/>Information Technology & Electronic Engineering, 16(4):272-282. +<br/>[doi:10.1631/FITEE.1400209] +<br/>Using Kinect for real-time emotion +<br/>recognition via facial expressions +<br/>Key words: Kinect, Emotion recognition, Facial expression, Real-time +<br/>classification, Fusion algorithm, Support vector machine (SVM) +<br/> ORCID: http://orcid.org/0000-0002-5021-9057 +<br/>Front Inform Technol & Electron Eng </td></tr><tr><td>004a1bb1a2c93b4f379468cca6b6cfc6d8746cc4</td><td>Balanced k-Means and Min-Cut Clustering +</td></tr><tr><td>00d94b35ffd6cabfb70b9a1d220b6823ae9154ee</td><td>Discriminative Bayesian Dictionary Learning +<br/>for Classification +</td></tr><tr><td>006f283a50d325840433f4cf6d15876d475bba77</td><td>756 +<br/>Preserving Structure in Model-Free Tracking +</td></tr><tr><td>0059b3dfc7056f26de1eabaafd1ad542e34c2c2e</td><td></td></tr><tr><td>6e198f6cc4199e1c4173944e3df6f39a302cf787</td><td>MORPH-II: Inconsistencies and Cleaning Whitepaper +<br/>NSF-REU Site at UNC Wilmington, Summer 2017 +</td></tr><tr><td>6eaf446dec00536858548fe7cc66025b70ce20eb</td><td></td></tr><tr><td>6e91be2ad74cf7c5969314b2327b513532b1be09</td><td>Dimensionality Reduction with Subspace Structure +<br/>Preservation +<br/>Department of Computer Science +<br/>SUNY Buffalo +<br/>Buffalo, NY 14260 +</td></tr><tr><td>6eba25166fe461dc388805cc2452d49f5d1cdadd</td><td>Pages 122.1-122.12 +<br/>DOI: https://dx.doi.org/10.5244/C.30.122 +</td></tr><tr><td>6e8a81d452a91f5231443ac83e4c0a0db4579974</td><td>Illumination robust face representation based on intrinsic geometrical +<br/>information +<br/>Soyel, H; Ozmen, B; McOwan, PW +<br/>This is a pre-copyedited, author-produced PDF of an article accepted for publication in IET +<br/>Conference on Image Processing (IPR 2012). The version of record is available +<br/>http://ieeexplore.ieee.org/document/6290632/?arnumber=6290632&tag=1 +<br/>For additional information about this publication click this link. +<br/>http://qmro.qmul.ac.uk/xmlui/handle/123456789/16147 +<br/>Information about this research object was correct at the time of download; we occasionally +<br/>make corrections to records, please therefore check the published record when citing. For +</td></tr><tr><td>6ecd4025b7b5f4894c990614a9a65e3a1ac347b2</td><td>International Journal on Recent and Innovation Trends in Computing and Communication +<br/> +<br/> ISSN: 2321-8169 +<br/>Volume: 2 Issue: 5 +<br/> 1275– 1281 +<br/>_______________________________________________________________________________________________ +<br/>Automatic Naming of Character using Video Streaming for Face +<br/>Recognition with Graph Matching +<br/>Nivedita.R.Pandey +<br/>Ranjan.P.Dahake +<br/>PG Student at MET’s IOE Bhujbal Knowledge City, +<br/>PG Student at MET’s IOE Bhujbal Knowledge City, +<br/>Nasik, Maharashtra, India, +<br/>Nasik, Maharashtra, India, +</td></tr><tr><td>6eaeac9ae2a1697fa0aa8e394edc64f32762f578</td><td></td></tr><tr><td>6ee2ea416382d659a0dddc7a88fc093accc2f8ee</td><td></td></tr><tr><td>6e3a181bf388dd503c83dc324561701b19d37df1</td><td>Finding a low-rank basis in a matrix subspace +<br/>Andr´e Uschmajew +</td></tr><tr><td>6e8c3b7d25e6530a631ea01fbbb93ac1e8b69d2f</td><td>Deep Episodic Memory: Encoding, Recalling, and Predicting +<br/>Episodic Experiences for Robot Action Execution +</td></tr><tr><td>6e911227e893d0eecb363015754824bf4366bdb7</td><td>Wasserstein Divergence for GANs +<br/>1 Computer Vision Lab, ETH Zurich, Switzerland +<br/>2 VISICS, KU Leuven, Belgium +</td></tr><tr><td>6ee8a94ccba10062172e5b31ee097c846821a822</td><td>Submitted 3/13; Revised 10/13; Published 12/13 +<br/>How to Solve Classification and Regression Problems on +<br/>High-Dimensional Data with a Supervised +<br/>Extension of Slow Feature Analysis +<br/>Institut f¨ur Neuroinformatik +<br/>Ruhr-Universit¨at Bochum +<br/>Bochum D-44801, Germany +<br/>Editor: David Dunson +</td></tr><tr><td>6e379f2d34e14efd85ae51875a4fa7d7ae63a662</td><td>A NEW MULTI-MODAL BIOMETRIC SYSTEM +<br/>BASED ON FINGERPRINT AND FINGER +<br/>VEIN RECOGNITION +<br/>Master's Thesis +<br/>Department of Software Engineering +<br/>JULY-2014 +<br/>I +</td></tr><tr><td>6e1802874ead801a7e1072aa870681aa2f555f35</td><td>1424407281/07/$20.00 ©2007 IEEE +<br/>I 629 +<br/>ICASSP 2007 +</td></tr><tr><td>6ed22b934e382c6f72402747d51aa50994cfd97b</td><td>Customized Expression Recognition for Performance-Driven +<br/>Cutout Character Animation +<br/>†NEC Laboratories America +<br/>‡Snapchat +</td></tr><tr><td>6e93fd7400585f5df57b5343699cb7cda20cfcc2</td><td>http://journalofvision.org/9/2/22/ +<br/>Comparing a novel model based on the transferable +<br/>belief model with humans during the recognition of +<br/>partially occluded facial expressions +<br/>Département de Psychologie, Université de Montréal, +<br/>Canada +<br/>Département de Psychologie, Université de Montréal, +<br/>Canada +<br/>Département de Psychologie, Université de Montréal, +<br/>Canada +<br/>Humans recognize basic facial expressions effortlessly. Yet, despite a considerable amount of research, this task remains +<br/>elusive for computer vision systems. Here, we compared the behavior of one of the best computer models of facial +<br/>expression recognition (Z. Hammal, L. Couvreur, A. Caplier, & M. Rombaut, 2007) with the behavior of human observers +<br/>during the M. Smith, G. Cottrell, F. Gosselin, and P. G. Schyns (2005) facial expression recognition task performed on +<br/>stimuli randomly sampled using Gaussian apertures. The modelVwhich we had to significantly modify in order to give the +<br/>ability to deal with partially occluded stimuliVclassifies the six basic facial expressions (Happiness, Fear, Sadness, +<br/>Surprise, Anger, and Disgust) plus Neutral from static images based on the permanent facial feature deformations and the +<br/>Transferable Belief Model (TBM). Three simulations demonstrated the suitability of the TBM-based model to deal with +<br/>partially occluded facial parts and revealed the differences between the facial information used by humans and by the +<br/>model. This opens promising perspectives for the future development of the model. +<br/>Keywords: facial features behavior, facial expressions classification, Transferable Belief Model, Bubbles +<br/>Citation: Hammal, Z., Arguin, M., & Gosselin, F. (2009). Comparing a novel model based on the transferable belief +<br/>http://journalofvision.org/9/2/22/, doi:10.1167/9.2.22. +<br/>Introduction +<br/>Facial expressions communicate information from +<br/>which we can quickly infer the state of mind of our peers +<br/>and adjust our behavior accordingly (Darwin, 1872). To +<br/>illustrate, take a person like patient SM with complete +<br/>bilateral damage to the amygdala nuclei that prevents her +<br/>from recognizing facial expressions of fear. SM would be +<br/>incapable of interpreting the fearful expression on the face +<br/>of a bystander, who has encountered a furious Grizzly +<br/>bear, as a sign of potential +<br/>threat (Adolphs, Tranel, +<br/>Damasio, & Damasio, 1994). +<br/>Facial expressions are typically arranged into six +<br/>universally recognized basic categories Happiness, Sur- +<br/>prise, Disgust, Anger, Sadness, and Fear that are similarly +<br/>expressed across different backgrounds and cultures +<br/>(Cohn, 2006; Ekman, 1999; Izard, 1971, 1994). Facial +<br/>expressions result +<br/>from the precisely choreographed +<br/>deformation of facial features, which are often described +<br/>using the 46 Action Units (AUs; Ekman & Friesen, +<br/>1978). +<br/>Facial expression recognition and computer +<br/>vision +<br/>The study of human facial expressions has an impact in +<br/>several areas of life such as art, social interaction, cognitive +<br/>science, medicine, security, affective computing, and +<br/>human-computer interaction (HCI). An automatic facial +<br/>expressions classification system may contribute signifi- +<br/>cantly to the development of all these disciplines. However, +<br/>the development of such a system constitutes a significant +<br/>challenge because of the many constraints that are imposed +<br/>by its application in a real-world context (Pantic & Bartlett, +<br/>2007; Pantic & Patras, 2006). In particular, such systems +<br/>need to provide great accuracy and robustness without +<br/>demanding too many interventions from the user. +<br/>There have been major advances in computer vision +<br/>over the past 15 years for the recognition of the six basic +<br/>facial expressions (for reviews, see Fasel & Luettin, 2003; +<br/>Pantic & Rothkrantz, 2000b). The main approaches can be +<br/>divided in two classes: Model-based and fiducial points +<br/>approaches. The model-based approach requires the +<br/>design of a deterministic physical model that can represent +<br/>doi: 10.1167/9.2.22 +<br/>Received January 28, 2008; published February 26, 2009 +<br/>ISSN 1534-7362 * ARVO +</td></tr><tr><td>6e12ba518816cbc2d987200c461dc907fd19f533</td><td></td></tr><tr><td>9ab463d117219ed51f602ff0ddbd3414217e3166</td><td>Weighted Transmedia +<br/>Relevance Feedback for +<br/>Image Retrieval and +<br/>Auto-annotation +<br/>TECHNICAL +<br/>REPORT +<br/>N° 0415 +<br/>December 2011 +<br/>Project-Teams LEAR - INRIA +<br/>and TVPA - XRCE +</td></tr><tr><td>9ac82909d76b4c902e5dde5838130de6ce838c16</td><td>Recognizing Facial Expressions Automatically +<br/>from Video +<br/>1 Introduction +<br/>Facial expressions, resulting from movements of the facial muscles, are the face +<br/>changes in response to a person’s internal emotional states, intentions, or social +<br/>communications. There is a considerable history associated with the study on fa- +<br/>cial expressions. Darwin (1872) was the first to describe in details the specific fa- +<br/>cial expressions associated with emotions in animals and humans, who argued that +<br/>all mammals show emotions reliably in their faces. Since that, facial expression +<br/>analysis has been a area of great research interest for behavioral scientists (Ekman, +<br/>Friesen, and Hager, 2002). Psychological studies (Mehrabian, 1968; Ambady and +<br/>Rosenthal, 1992) suggest that facial expressions, as the main mode for non-verbal +<br/>communication, play a vital role in human face-to-face communication. For illus- +<br/>tration, we show some examples of facial expressions in Fig. 1. +<br/>Computer recognition of facial expressions has many important applications in +<br/>intelligent human-computer interaction, computer animation, surveillance and se- +<br/>curity, medical diagnosis, law enforcement, and awareness systems (Shan, 2007). +<br/>Therefore, it has been an active research topic in multiple disciplines such as psy- +<br/>chology, cognitive science, human-computer interaction, and pattern recognition. +<br/>Meanwhile, as a promising unobtrusive solution, automatic facial expression analy- +<br/>sis from video or images has received much attention in last two decades (Pantic and +<br/>Rothkrantz, 2000a; Fasel and Luettin, 2003; Tian, Kanade, and Cohn, 2005; Pantic +<br/>and Bartlett, 2007). +<br/>This chapter introduces recent advances in computer recognition of facial expres- +<br/>sions. Firstly, we describe the problem space, which includes multiple dimensions: +<br/>level of description, static versus dynamic expression, facial feature extraction and +</td></tr><tr><td>9ac15845defcd0d6b611ecd609c740d41f0c341d</td><td>Copyright +<br/>by +<br/>2011 +</td></tr><tr><td>9af1cf562377b307580ca214ecd2c556e20df000</td><td>Feb. 28 +<br/> International Journal of Advanced Studies in Computer Science and Engineering +<br/>IJASCSE, Volume 4, Issue 2, 2015 +<br/> Video-Based Facial Expression Recognition +<br/>Using Local Directional Binary Pattern +<br/>Electrical Engineering Dept., AmirKabir Univarsity of Technology +<br/>Tehran, Iran +</td></tr><tr><td>9a23a0402ae68cc6ea2fe0092b6ec2d40f667adb</td><td>High-Resolution Image Synthesis and Semantic Manipulation with Conditional GANs +<br/>1NVIDIA Corporation +<br/>2UC Berkeley +<br/>Figure 1: We propose a generative adversarial framework for synthesizing 2048 × 1024 images from semantic label maps +<br/>(lower left corner in (a)). Compared to previous work [5], our results express more natural textures and details. (b) We can +<br/>change labels in the original label map to create new scenes, like replacing trees with buildings. (c) Our framework also +<br/>allows a user to edit the appearance of individual objects in the scene, e.g. changing the color of a car or the texture of a road. +<br/>Please visit our website for more side-by-side comparisons as well as interactive editing demos. +</td></tr><tr><td>9a7858eda9b40b16002c6003b6db19828f94a6c6</td><td>MOONEY FACE CLASSIFICATION AND PREDICTION BY LEARNING ACROSS TONE +<br/>(cid:63) UC Berkeley / †ICSI +</td></tr><tr><td>9a276c72acdb83660557489114a494b86a39f6ff</td><td>Emotion Classification through Lower Facial Expressions using Adaptive +<br/>Support Vector Machines +<br/>Department of Information Technology, Faculty of Industrial Technology and Management, +</td></tr><tr><td>9a42c519f0aaa68debbe9df00b090ca446d25bc4</td><td>Face Recognition via Centralized Coordinate +<br/>Learning +</td></tr><tr><td>9aad8e52aff12bd822f0011e6ef85dfc22fe8466</td><td>Temporal-Spatial Mapping for Action Recognition +</td></tr><tr><td>36b40c75a3e53c633c4afb5a9309d10e12c292c7</td><td></td></tr><tr><td>3646b42511a6a0df5470408bc9a7a69bb3c5d742</td><td>International Journal of Computer Applications (0975 – 8887) +<br/>Applications of Computers and Electronics for the Welfare of Rural Masses (ACEWRM) 2015 +<br/>Detection of Facial Parts based on ABLATA +<br/>Technical Campus, Bhilai +<br/>Vikas Singh +<br/>Technical Campus, Bhilai +<br/>Abha Choubey +<br/>Technical Campus, Bhilai +</td></tr><tr><td>365f67fe670bf55dc9ccdcd6888115264b2a2c56</td><td></td></tr><tr><td>36fe39ed69a5c7ff9650fd5f4fe950b5880760b0</td><td>Tracking von Gesichtsmimik +<br/>mit Hilfe von Gitterstrukturen +<br/>zur Klassifikation von schmerzrelevanten Action +<br/>Units +<br/>1Fraunhofer-Institut f¨ur Integrierte Schaltungen IIS, Erlangen, +<br/>2Otto-Friedrich-Universit¨at Bamberg, 3Universit¨atsklinkum Erlangen +<br/>Kurzfassung. In der Schmerzforschung werden schmerzrelevante Mi- +<br/>mikbewegungen von Probanden mittels des Facial Action Coding System +<br/>klassifiziert. Die manuelle Klassifikation hierbei ist aufw¨andig und eine +<br/>automatische (Vor-)klassifikation k¨onnte den diagnostischen Wert dieser +<br/>Analysen erh¨ohen sowie den klinischen Workflow unterst¨utzen. Der hier +<br/>vorgestellte regelbasierte Ansatz erm¨oglicht eine automatische Klassifika- +<br/>tion ohne große Trainingsmengen vorklassifizierter Daten. Das Verfahren +<br/>erkennt und verfolgt Mimikbewegungen, unterst¨utzt durch ein Gitter, +<br/>und ordnet diese Bewegungen bestimmten Gesichtsarealen zu. Mit die- +<br/>sem Wissen kann aus den Bewegungen auf die zugeh¨origen Action Units +<br/>geschlossen werden. +<br/>1 Einleitung +<br/>Menschliche Empfindungen wie Emotionen oder Schmerz l¨osen spezifische Mu- +<br/>ster von Kontraktionen der Gesichtsmuskulatur aus, die Grundlage dessen sind, +<br/>was wir Mimik nennen. Aus der Beobachtung der Mimik kann wiederum auf +<br/>menschliche Empfindungen r¨uckgeschlossen werden. Im Rahmen der Schmerz- +<br/>forschung werden Videoaufnahmen von Probanden hinsichtlich des mimischen +<br/>Schmerzausdrucks analysiert. Zur Beschreibung des mimischen Ausdrucks und +<br/>dessen Ver¨anderungen wird das Facial Action Coding System (FACS) [1] verwen- +<br/>det, das anatomisch begr¨undet, kleinste sichtbare Muskelbewegungen im Gesicht +<br/>beschreibt und als einzelne Action Units (AUs) kategorisiert. Eine Vielzahl von +<br/>Untersuchungen hat gezeigt, dass spezifische Muster von Action Units auftre- +<br/>ten, wenn Probanden Schmerzen angeben [2]. Die manuelle Klassifikation und +<br/>Markierung der Action Units von Probanden in Videosequenzen bedarf einer +<br/>langwierigen Beobachtung durch ausgebildete FACS-Coder. Eine automatische +<br/>(Vor-)klassifikation kann hierbei den klinischen Workflow unterst¨utzen und dieses +<br/>Verfahren zum brauchbaren diagnostischen Instrument machen. Bisher realisier- +<br/>te Ans¨atze zum Erkennen von Gesichtsausdr¨ucken basieren auf der Klassifikation +</td></tr><tr><td>36ce0b68a01b4c96af6ad8c26e55e5a30446f360</td><td>Multimed Tools Appl +<br/>DOI 10.1007/s11042-014-2322-6 +<br/>Facial expression recognition based on a mlp neural +<br/>network using constructive training algorithm +<br/>Received: 5 February 2014 / Revised: 22 August 2014 / Accepted: 13 October 2014 +<br/>© Springer Science+Business Media New York 2014 +</td></tr><tr><td>3674f3597bbca3ce05e4423611d871d09882043b</td><td>ISSN 1796-2048 +<br/>Volume 7, Number 4, August 2012 +<br/>Contents +<br/>Special Issue: Multimedia Contents Security in Social Networks Applications +<br/>Guest Editors: Zhiyong Zhang and Muthucumaru Maheswaran +<br/>Guest Editorial +<br/>Zhiyong Zhang and Muthucumaru Maheswaran +<br/>SPECIAL ISSUE PAPERS +<br/>DRTEMBB: Dynamic Recommendation Trust Evaluation Model Based on Bidding +<br/>Gang Wang and Xiao-lin Gui +<br/>Block-Based Parallel Intra Prediction Scheme for HEVC +<br/>Jie Jiang, Baolong, Wei Mo, and Kefeng Fan +<br/>Optimized LSB Matching Steganography Based on Fisher Information +<br/>Yi-feng Sun, Dan-mei Niu, Guang-ming Tang, and Zhan-zhan Gao +<br/>A Novel Robust Zero-Watermarking Scheme Based on Discrete Wavelet Transform +<br/>Yu Yang, Min Lei, Huaqun Liu, Yajian Zhou, and Qun Luo +<br/>Stego Key Estimation in LSB Steganography +<br/>Jing Liu and Guangming Tang +<br/>REGULAR PAPERS +<br/>Facial Expression Spacial Charts for Describing Dynamic Diversity of Facial Expressions +<br/>277 +<br/>279 +<br/>289 +<br/>295 +<br/>303 +<br/>309 +<br/>314 +</td></tr><tr><td>362a70b6e7d55a777feb7b9fc8bc4d40a57cde8c</td><td>978-1-4799-9988-0/16/$31.00 ©2016 IEEE +<br/>2792 +<br/>ICASSP 2016 +</td></tr><tr><td>366d20f8fd25b4fe4f7dc95068abc6c6cabe1194</td><td></td></tr><tr><td>3630324c2af04fd90f8668f9ee9709604fe980fd</td><td>This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCSVT.2016.2607345, IEEE +<br/>Transactions on Circuits and Systems for Video Technology +<br/>Image Classification with Tailored Fine-Grained +<br/>Dictionaries +</td></tr><tr><td>362ba8317aba71c78dafca023be60fb71320381d</td><td></td></tr><tr><td>36cf96fe11a2c1ea4d999a7f86ffef6eea7b5958</td><td>RGB-D Face Recognition with Texture and +<br/>Attribute Features +<br/>Member, IEEE +</td></tr><tr><td>36018404263b9bb44d1fddaddd9ee9af9d46e560</td><td>OCCLUDED FACE RECOGNITION BY USING GABOR +<br/>FEATURES +<br/>1 Department of Electrical And Electronics Engineering, METU, Ankara, Turkey +<br/>2 7h%ł7$.(cid:3)%ł/7(1(cid:15)(cid:3)$QNDUD(cid:15)(cid:3)7XUNH\ +</td></tr><tr><td>5c4ce36063dd3496a5926afd301e562899ff53ea</td><td></td></tr><tr><td>5c2a7518fb26a37139cebff76753d83e4da25159</td><td></td></tr><tr><td>5c2e264d6ac253693469bd190f323622c457ca05</td><td>978-1-4799-2341-0/13/$31.00 ©2013 IEEE +<br/>4367 +<br/>ICIP 2013 +</td></tr><tr><td>5c473cfda1d7c384724fbb139dfe8cb39f79f626</td><td></td></tr><tr><td>5c5e1f367e8768a9fb0f1b2f9dbfa060a22e75c0</td><td>2132 +<br/>Reference Face Graph for Face Recognition +</td></tr><tr><td>5c35ac04260e281141b3aaa7bbb147032c887f0c</td><td>Face Detection and Tracking Control with Omni Car +<br/>CS 231A Final Report +<br/>June 31, 2016 +</td></tr><tr><td>5c717afc5a9a8ccb1767d87b79851de8d3016294</td><td>978-1-4673-0046-9/12/$26.00 ©2012 IEEE +<br/>1845 +<br/>ICASSP 2012 +</td></tr><tr><td>0952ac6ce94c98049d518d29c18d136b1f04b0c0</td><td></td></tr><tr><td>09137e3c267a3414314d1e7e4b0e3a4cae801f45</td><td>Noname manuscript No. +<br/>(will be inserted by the editor) +<br/>Two Birds with One Stone: Transforming and Generating +<br/>Facial Images with Iterative GAN +<br/>Received: date / Accepted: date +</td></tr><tr><td>09926ed62511c340f4540b5bc53cf2480e8063f8</td><td>Action Tubelet Detector for Spatio-Temporal Action Localization +</td></tr><tr><td>09718bf335b926907ded5cb4c94784fd20e5ccd8</td><td>875 +<br/>Recognizing Partially Occluded, Expression Variant +<br/>Faces From Single Training Image per Person +<br/>With SOM and Soft k-NN Ensemble +</td></tr><tr><td>0903bb001c263e3c9a40f430116d1e629eaa616f</td><td>CVPR +<br/>#987 +<br/>000 +<br/>001 +<br/>002 +<br/>003 +<br/>004 +<br/>005 +<br/>006 +<br/>007 +<br/>008 +<br/>009 +<br/>010 +<br/>011 +<br/>012 +<br/>013 +<br/>014 +<br/>015 +<br/>016 +<br/>017 +<br/>018 +<br/>019 +<br/>020 +<br/>021 +<br/>022 +<br/>023 +<br/>024 +<br/>025 +<br/>026 +<br/>027 +<br/>028 +<br/>029 +<br/>030 +<br/>031 +<br/>032 +<br/>033 +<br/>034 +<br/>035 +<br/>036 +<br/>037 +<br/>038 +<br/>039 +<br/>040 +<br/>041 +<br/>042 +<br/>043 +<br/>044 +<br/>045 +<br/>046 +<br/>047 +<br/>048 +<br/>049 +<br/>050 +<br/>051 +<br/>052 +<br/>053 +<br/>CVPR 2009 Submission #987. CONFIDENTIAL REVIEW COPY. DO NOT DISTRIBUTE. +<br/>An Empirical Study of Context in Object Detection +<br/>Anonymous CVPR submission +<br/>Paper ID 987 +</td></tr><tr><td>09df62fd17d3d833ea6b5a52a232fc052d4da3f5</td><td>ISSN: 1405-5546 +<br/>Instituto Politécnico Nacional +<br/>México +<br/> +<br/>Rivas Araiza, Edgar A.; Mendiola Santibañez, Jorge D.; Herrera Ruiz, Gilberto; González Gutiérrez, +<br/>Carlos A.; Trejo Perea, Mario; Ríos Moreno, G. J. +<br/>Mejora de Contraste y Compensación en Cambios de la Iluminación +<br/>Instituto Politécnico Nacional +<br/>Distrito Federal, México +<br/>Disponible en: http://www.redalyc.org/articulo.oa?id=61509703 +<br/> Cómo citar el artículo +<br/> Número completo +<br/> Más información del artículo +<br/> Página de la revista en redalyc.org +<br/>Sistema de Información Científica +<br/>Red de Revistas Científicas de América Latina, el Caribe, España y Portugal +<br/>Proyecto académico sin fines de lucro, desarrollado bajo la iniciativa de acceso abierto +</td></tr><tr><td>097104fc731a15fad07479f4f2c4be2e071054a2</td><td></td></tr><tr><td>09f853ce12f7361c4b50c494df7ce3b9fad1d221</td><td>myjournal manuscript No. +<br/>(will be inserted by the editor) +<br/>Random forests for real time 3D face analysis +<br/>Received: date / Accepted: date +</td></tr><tr><td>09111da0aedb231c8484601444296c50ca0b5388</td><td></td></tr><tr><td>09750c9bbb074bbc4eb66586b20822d1812cdb20</td><td>978-1-4673-0046-9/12/$26.00 ©2012 IEEE +<br/>1385 +<br/>ICASSP 2012 +</td></tr><tr><td>097f674aa9e91135151c480734dda54af5bc4240</td><td>Proc. VIIth Digital Image Computing: Techniques and Applications, Sun C., Talbot H., Ourselin S. and Adriaansen T. (Eds.), 10-12 Dec. 2003, Sydney +<br/>Face Recognition Based on Multiple Region Features +<br/>CSIRO Telecommunications & Industrial Physics +<br/>Australia +<br/>Tel: 612 9372 4104, Fax: 612 9372 4411, Email: +</td></tr><tr><td>5d485501f9c2030ab33f97972aa7585d3a0d59a7</td><td></td></tr><tr><td>5de5848dc3fc35e40420ffec70a407e4770e3a8d</td><td>WebVision Database: Visual Learning and Understanding from Web Data +<br/>1 Computer Vision Laboratory, ETH Zurich +<br/>2 Google Switzerland +</td></tr><tr><td>5da139fc43216c86d779938d1c219b950dd82a4c</td><td>1-4244-1437-7/07/$20.00 ©2007 IEEE +<br/>II - 205 +<br/>ICIP 2007 +</td></tr><tr><td>5dc056fe911a3e34a932513abe637076250d96da</td><td></td></tr><tr><td>5d233e6f23b1c306cf62af49ce66faac2078f967</td><td>RESEARCH ARTICLE +<br/>Optimal Geometrical Set for Automated +<br/>Marker Placement to Virtualized Real-Time +<br/>Facial Emotions +<br/>School of Mechatronic Engineering, Universiti Malaysia Perlis, 02600, Ulu Pauh, Arau, Perlis, West Malaysia +</td></tr><tr><td>5d7f8eb73b6a84eb1d27d1138965eb7aef7ba5cf</td><td>Robust Registration of Dynamic Facial Sequences +</td></tr><tr><td>5dcf78de4d3d867d0fd4a3105f0defae2234b9cb</td><td></td></tr><tr><td>5db4fe0ce9e9227042144758cf6c4c2de2042435</td><td>INTERNATIONAL JOURNAL OF ELECTRICAL AND ELECTRONIC SYSTEMS RESEARCH, VOL.3, JUNE 2010 +<br/>Recognition of Facial Expression Using Haar +<br/>Wavelet Transform +<br/>for +<br/>paper +<br/>features +<br/>investigates +<br/> +</td></tr><tr><td>5d5cd6fa5c41eb9d3d2bab3359b3e5eb60ae194e</td><td>Face Recognition Algorithms +<br/>June 16, 2010 +<br/>Ion Marqu´es +<br/>Supervisor: +<br/>Manuel Gra˜na +</td></tr><tr><td>5d09d5257139b563bd3149cfd5e6f9eae3c34776</td><td>Optics Communications 338 (2015) 77–89 +<br/>Contents lists available at ScienceDirect +<br/>Optics Communications +<br/>journal homepage: www.elsevier.com/locate/optcom +<br/>Pattern recognition with composite correlation filters designed with +<br/>multi-objective combinatorial optimization +<br/>a Instituto Politécnico Nacional – CITEDI, Ave. del Parque 1310, Mesade Otay, Tijuana B.C. 22510, México +<br/>b Department of Computer Science, CICESE, Carretera Ensenada-Tijuana 3918, Ensenada B.C. 22860, México +<br/>c Instituto Tecnológico de Tijuana, Blvd. Industrial y Ave. ITR TijuanaS/N, Mesa de Otay, Tijuana B.C. 22500, México +<br/>d National Ignition Facility, Lawrence Livermore National Laboratory, Livermore, CA 94551, USA +<br/>a r t i c l e i n f o +<br/>a b s t r a c t +<br/>Article history: +<br/>Received 12 July 2014 +<br/>Accepted 16 November 2014 +<br/>Available online 23 October 2014 +<br/>Keywords: +<br/>Object recognition +<br/>Composite correlation filters +<br/>Multi-objective evolutionary algorithm +<br/>Combinatorial optimization +<br/>Composite correlation filters are used for solving a wide variety of pattern recognition problems. These +<br/>filters are given by a combination of several training templates chosen by a designer in an ad hoc manner. +<br/>In this work, we present a new approach for the design of composite filters based on multi-objective +<br/>combinatorial optimization. Given a vast search space of training templates, an iterative algorithm is used +<br/>to synthesize a filter with an optimized performance in terms of several competing criteria. Moreover, by +<br/>employing a suggested binary-search procedure a filter bank with a minimum number of filters can be +<br/>constructed, for a prespecified trade-off of performance metrics. Computer simulation results obtained +<br/>with the proposed method in recognizing geometrically distorted versions of a target in cluttered and +<br/>noisy scenes are discussed and compared in terms of recognition performance and complexity with +<br/>existing state-of-the-art filters. +<br/>& Elsevier B.V. All rights reserved. +<br/>1. +<br/>Introduction +<br/>Nowadays, object recognition receives much research interest +<br/>due to its high impact in real-life activities, such as robotics, bio- +<br/>metrics, and target tracking [1,2]. Object recognition consists in +<br/>solving two essential tasks: detection of a target within an ob- +<br/>served scene and determination of the exact position of the de- +<br/>tected object. Different approaches can be utilized to address these +<br/>tasks, that is feature-based methods [3–6] and template matching +<br/>algorithms [7,8]. In feature-based methods the observed scene is +<br/>processed to extract relevant features of potential targets within +<br/>the scene. Next, the extracted features are processed and analyzed +<br/>to make decisions. Feature-based methods yield good results in +<br/>many applications. However, they depend on several subjective +<br/>decisions which often require optimization [9,10]. On the other +<br/>hand, correlation filtering is a template matching processing. In +<br/>this approach, the coordinates of the maximum of the filter output +<br/>are taken as estimates of the target coordinates in the observed +<br/>scene. Correlation filters possess a good mathematical basis and +<br/>they can be implemented by exploiting massive parallelism either +<br/>in hybrid opto-digital correlators [11,12] or in high-performance +<br/>n Corresponding author. Tel.: þ52 664 623 1344x82856. +<br/>http://dx.doi.org/10.1016/j.optcom.2014.10.038 +<br/>0030-4018/& Elsevier B.V. All rights reserved. +<br/>hardware such as graphics processing units (GPUs) [13] or field +<br/>programmable gate arrays (FPGAs) [14] at high rate. Additionally, +<br/>these filters are capable to reliably recognize a target in highly +<br/>cluttered and noisy environments [8,15,16]. Moreover, they are +<br/>able to estimate very accurately the position of the target within +<br/>the scene [17]. Correlation filters are usually designed by a opti- +<br/>mization of various criteria [18,19]. The filters can be broadly +<br/>classified in to two main categories: analytical and composite fil- +<br/>ters. Analytical filters optimize a performance criterion using +<br/>mathematical models of signals and noise [20,21]. Composite fil- +<br/>ters are constructed by combination of several training templates, +<br/>each of them representing an expected target view in the observed +<br/>scene [22,21]. In practice, composite filters are effective for real- +<br/>life degradations of targets such as rotations and scaling. Compo- +<br/>site filters are synthesized by means of a supervised training +<br/>process. Thus, the performance of the filters highly depends on a +<br/>proper selection of image templates used for training [20,23]. +<br/>Normally, the training templates are chosen by a designer in an ad +<br/>hoc manner. Such a subjective procedure is not optimal. In addi- +<br/>tion, Kumar and Pochavsky [24] showed that the signal to noise +<br/>ratio of a composite filter gradually reduces when the number of +<br/>training templates increases. In order to synthesize composite +<br/>filters with improved performance in terms of several competing +<br/>metrics, a search and optimization strategy is required to auto- +<br/>matically choose the set of training templates. +</td></tr><tr><td>5d01283474b73a46d80745ad0cc0c4da14aae194</td><td></td></tr><tr><td>5d197c8cd34473eb6cde6b65ced1be82a3a1ed14</td><td><b>AFaceImageDatabaseforEvaluatingOut-of-FocusBlurQiHan,QiongLiandXiamuNiuHarbinInstituteofTechnologyChina1.IntroductionFacerecognitionisoneofthemostpopularresearchfieldsofcomputervisionandmachinelearning(Tores(2004);Zhaoetal.(2003)).Alongwithinvestigationoffacerecognitionalgorithmsandsystems,manyfaceimagedatabaseshavebeencollected(Gross(2005)).Facedatabasesareimportantfortheadvancementoftheresearchfield.Becauseofthenonrigidityandcomplex3Dstructureofface,manyfactorsinfluencetheperformanceoffacedetectionandrecognitionalgorithmssuchaspose,expression,age,brightness,contrast,noise,blurandetc.Someearlyfacedatabasesgatheredunderstrictlycontrolledenvironment(Belhumeuretal.(1997);Samaria&Harter(1994);Turk&Pentland(1991))onlyallowslightexpressionvariation.Toinvestigatetherelationshipsbetweenalgorithms’performanceandtheabovefactors,morefacedatabaseswithlargerscaleandvariouscharacterswerebuiltinthepastyears(Bailly-Bailliereetal.(2003);Flynnetal.(2003);Gaoetal.(2008);Georghiadesetal.(2001);Hallinan(1995);Phillipsetal.(2000);Simetal.(2003)).Forinstance,The"CAS-PEAL","FERET","CMUPIE",and"YaleB"databasesincludevariousposes(Gaoetal.(2008);Georghiadesetal.(2001);Phillipsetal.(2000);Simetal.(2003));The"HarvardRL","CMUPIE"and"YaleB"databasesinvolvemorethan40differentconditionsinillumination(Georghiadesetal.(2001);Hallinan(1995);Simetal.(2003));Andthe"BANCA",and"NDHID"databasescontainover10timesgathering(Bailly-Bailliereetal.(2003);Flynnetal.(2003)).Thesedatabaseshelpresearcherstoevaluateandimprovetheiralgorithmsaboutfacedetection,recognition,andotherpurposes.Blurisnotthemostimportantbutstillanotablefactoraffectingtheperformanceofabiometricsystem(Fronthaleretal.(2006);Zamanietal.(2007)).Themainreasonsleadingblurconsistinout-of-focusofcameraandmotionofobject,andtheout-of-focusblurismoresignificantintheapplicationenvironmentoffacerecognition(Eskicioglu&Fisher(1995);Kimetal.(1998);Tanakaetal.(2007);Yitzhaky&Kopeika(1996)).Toinvestigatetheinfluenceofbluronafacerecognitionsystem,afaceimagedatabasewithdifferentconditionsofclarityandefficientblurevaluatingalgorithmsareneeded.Thischapterintroducesanewfacedatabasebuiltforthepurposeofblurevaluation.Theapplicationenvironmentsoffacerecognitionareanalyzedfirstly,thenaimagegatheringschemeisdesigned.Twotypicalgatheringfacilitiesareusedandthefocusstatusaredividedinto11steps.Further,theblurassessmentalgorithmsaresummarizedandthecomparisonbetweenthemisraisedonthevarious-claritydatabase.The7www.intechopen.com</b></td></tr><tr><td>31aa20911cc7a2b556e7d273f0bdd5a2f0671e0a</td><td></td></tr><tr><td>31b05f65405534a696a847dd19c621b7b8588263</td><td></td></tr><tr><td>318e7e6daa0a799c83a9fdf7dd6bc0b3e89ab24a</td><td>Sparsity in Dynamics of Spontaneous +<br/>Subtle Emotions: Analysis & Application +</td></tr><tr><td>31c0968fb5f587918f1c49bf7fa51453b3e89cf7</td><td>Deep Transfer Learning for Person Re-identification +</td></tr><tr><td>31e57fa83ac60c03d884774d2b515813493977b9</td><td></td></tr><tr><td>316e67550fbf0ba54f103b5924e6537712f06bee</td><td>Multimodal semi-supervised learning +<br/>for image classification +<br/>LEAR team, INRIA Grenoble, France +</td></tr><tr><td>31ef5419e026ef57ff20de537d82fe3cfa9ee741</td><td>Facial Expression Analysis Based on +<br/>High Dimensional Binary Features +<br/>´Ecole Polytechique de Montr´eal, Universit´e de Montr´eal, Montr´eal, Canada +</td></tr><tr><td>31b58ced31f22eab10bd3ee2d9174e7c14c27c01</td><td></td></tr><tr><td>31ace8c9d0e4550a233b904a0e2aabefcc90b0e3</td><td>Learning Deep Face Representation +<br/>Megvii Inc. +<br/>Megvii Inc. +<br/>Megvii Inc. +<br/>Megvii Inc. +<br/>Megvii Inc. +</td></tr><tr><td>312afff739d1e0fcd3410adf78be1c66b3480396</td><td></td></tr><tr><td>31bb49ba7df94b88add9e3c2db72a4a98927bb05</td><td></td></tr><tr><td>91811203c2511e919b047ebc86edad87d985a4fa</td><td>Expression Subspace Projection for Face +<br/>Recognition from Single Sample per Person +</td></tr><tr><td>91e57667b6fad7a996b24367119f4b22b6892eca</td><td>Probabilistic Corner Detection for Facial Feature +<br/>Extraction +<br/>Article +<br/>Accepted version +<br/>E. Ardizzone, M. La Cascia, M. Morana +<br/>In Lecture Notes in Computer Science Volume 5716, 2009 +<br/>It is advisable to refer to the publisher's version if you intend to cite +<br/>from the work. +<br/>Publisher: Springer +<br/>http://link.springer.com/content/pdf/10.1007%2F978-3- +<br/>642-04146-4_50.pdf +</td></tr><tr><td>91883dabc11245e393786d85941fb99a6248c1fb</td><td></td></tr><tr><td>917bea27af1846b649e2bced624e8df1d9b79d6f</td><td>Ultra Power-Efficient CNN Domain Specific Accelerator with 9.3TOPS/Watt for +<br/>Mobile and Embedded Applications +<br/>Gyrfalcon Technology Inc. +<br/>1900 McCarthy Blvd. Milpitas, CA 95035 +</td></tr><tr><td>91b1a59b9e0e7f4db0828bf36654b84ba53b0557</td><td>This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI +<br/>> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < +<br/> +<br/>Simultaneous Hallucination and Recognition of +<br/>Low-Resolution Faces Based on Singular Value +<br/>Decomposition +<br/>(SVD) +<br/>for performing both +</td></tr><tr><td>911bef7465665d8b194b6b0370b2b2389dfda1a1</td><td>RANJAN, ROMERO, BLACK: LEARNING HUMAN OPTICAL FLOW +<br/>Learning Human Optical Flow +<br/>1 MPI for Intelligent Systems +<br/>Tübingen, Germany +<br/>2 Amazon Inc. +</td></tr><tr><td>91ead35d1d2ff2ea7cf35d15b14996471404f68d</td><td>Combining and Steganography of 3D Face Textures +</td></tr><tr><td>919d0e681c4ef687bf0b89fe7c0615221e9a1d30</td><td></td></tr><tr><td>912a6a97af390d009773452814a401e258b77640</td><td></td></tr><tr><td>91d513af1f667f64c9afc55ea1f45b0be7ba08d4</td><td>Automatic Face Image Quality Prediction +</td></tr><tr><td>918b72a47b7f378bde0ba29c908babf6dab6f833</td><td></td></tr><tr><td>91e58c39608c6eb97b314b0c581ddaf7daac075e</td><td>Pixel-wise Ear Detection with Convolutional +<br/>Encoder-Decoder Networks +</td></tr><tr><td>91d2fe6fdf180e8427c65ffb3d895bf9f0ec4fa0</td><td></td></tr><tr><td>9131c990fad219726eb38384976868b968ee9d9c</td><td>Deep Facial Expression Recognition: A Survey +</td></tr><tr><td>915d4a0fb523249ecbc88eb62cb150a60cf60fa0</td><td>Comparison of Feature Extraction Techniques in Automatic +<br/>Face Recognition Systems for Security Applications +<br/>S . Cruz-Llanas, J. Ortega-Garcia, E. Martinez-Torrico, J. Gonzalez-Rodriguez +<br/>Dpto. Ingenieria Audiovisual y Comunicaciones, EUIT Telecomunicacion, Univ. PolitCcnica de Madrid, Spain +<br/>http://www.atvs.diac.upm.es +</td></tr><tr><td>65b737e5cc4a565011a895c460ed8fd07b333600</td><td>Transfer Learning For Cross-Dataset Recognition: A Survey +<br/>This paper summarises and analyses the cross-dataset recognition transfer learning techniques with the +<br/>emphasis on what kinds of methods can be used when the available source and target data are presented +<br/>in different forms for boosting the target task. This paper for the first time summarises several transferring +<br/>criteria in details from the concept level, which are the key bases to guide what kind of knowledge to transfer +<br/>between datasets. In addition, a taxonomy of cross-dataset scenarios and problems is proposed according the +<br/>properties of data that define how different datasets are diverged, thereby review the recent advances on +<br/>each specific problem under different scenarios. Moreover, some real world applications and corresponding +<br/>commonly used benchmarks of cross-dataset recognition are reviewed. Lastly, several future directions are +<br/>identified. +<br/>Additional Key Words and Phrases: Cross-dataset, transfer learning, domain adaptation +<br/>1. INTRODUCTION +<br/>It has been explored how human would transfer learning in one context to another +<br/>similar context [Woodworth and Thorndike 1901; Perkins et al. 1992] in the field of +<br/>Psychology and Education. For example, learning to drive a car helps a person later +<br/>to learn more quickly to drive a truck, and learning mathematics prepares students to +<br/>study physics. The machine learning algorithms are mostly inspired by human brains. +<br/>However, most of them require a huge amount of training examples to learn a new +<br/>model from scratch and fail to apply knowledge learned from previous domains or +<br/>tasks. This may be due to that a basic assumption of statistical learning theory is +<br/>that the training and test data are drawn from the same distribution and belong to +<br/>the same task. Intuitively, learning from scratch is not realistic and practical, because +<br/>it violates how human learn things. In addition, manually labelling a large amount +<br/>of data for new domain or task is labour extensive, especially for the modern “data- +<br/>hungry” and “data-driven” learning techniques (i.e. deep learning). However, the big +<br/>data era provides a huge amount available data collected for other domains and tasks. +<br/>Hence, how to use the previously available data smartly for the current task with +<br/>scarce data will be beneficial for real world applications. +<br/>To reuse the previous knowledge for current tasks, the differences between old data +<br/>and new data need to be taken into account. Take the object recognition as an ex- +<br/>ample. As claimed by Torralba and Efros [2011], despite the great efforts of object +<br/>datasets creators, the datasets appear to have strong build-in bias caused by various +<br/>factors, such as selection bias, capture bias, category or label bias, and negative set +<br/>bias. This suggests that no matter how big the dataset is, it is impossible to cover +<br/>the complexity of the real visual world. Hence, the dataset bias needs to be consid- +<br/>ered before reusing data from previous datasets. Pan and Yang [2010] summarise that +<br/>the differences between different datasets can be caused by domain divergence (i.e. +<br/>distribution shift or feature space difference) or task divergence (i.e. conditional dis- +<br/>tribution shift or label space difference), or both. For example, in visual recognition, +<br/>the distributions between the previous and current data can be discrepant due to the +<br/>different environments, lighting, background, sensor types, resolutions, view angles, +<br/>and post-processing. Those external factors may cause the distribution divergence or +<br/>even feature space divergence between different domains. On the other hand, the task +<br/>divergence between current and previous data is also ubiquitous. For example, it is +<br/>highly possible that an animal species that we want to recognize have not been seen +<br/>ACM Journal Name, Vol. V, No. N, Article A, Publication date: January YYYY. +</td></tr><tr><td>6582f4ec2815d2106957215ca2fa298396dde274</td><td>JUNE 2007 +<br/>1005 +<br/>Discriminative Learning and Recognition +<br/>of Image Set Classes Using +<br/>Canonical Correlations +</td></tr><tr><td>655d9ba828eeff47c600240e0327c3102b9aba7c</td><td>IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART B: CYBERNETICS, VOL. 35, NO. 3, JUNE 2005 +<br/>489 +<br/>Kernel Pooled Local Subspaces for Classification +</td></tr><tr><td>656a59954de3c9fcf82ffcef926af6ade2f3fdb5</td><td>Convolutional Network Representation +<br/>for Visual Recognition +<br/>Doctoral Thesis +<br/>Stockholm, Sweden, 2017 +</td></tr><tr><td>656f05741c402ba43bb1b9a58bcc5f7ce2403d9a</td><td></td></tr><tr><td>65817963194702f059bae07eadbf6486f18f4a0a</td><td>http://dx.doi.org/10.1007/s11263-015-0814-0 +<br/>WhittleSearch: Interactive Image Search with Relative Attribute +<br/>Feedback +<br/>Received: date / Accepted: date +</td></tr><tr><td>6581c5b17db7006f4cc3575d04bfc6546854a785</td><td>Contextual Person Identification +<br/>in Multimedia Data +<br/>zur Erlangung des akademischen Grades eines +<br/>Doktors der Ingenieurwissenschaften +<br/>der Fakultät für Informatik +<br/>des Karlsruher Instituts für Technologie (KIT) +<br/>genehmigte +<br/>Dissertation +<br/>von +<br/>aus Erlangen +<br/>Tag der mündlichen Prüfung: +<br/>18. November 2014 +<br/>Hauptreferent: +<br/>Korreferent: +<br/>Prof. Dr. Rainer Stiefelhagen +<br/>Karlsruher Institut für Technologie +<br/>Prof. Dr. Gerhard Rigoll +<br/>Technische Universität München +<br/>KIT – Universität des Landes Baden-Württemberg und nationales Forschungszentrum in der Helmholtz-Gemeinschaft +<br/>www.kit.edu +</td></tr><tr><td>65babb10e727382b31ca5479b452ee725917c739</td><td>Label Distribution Learning +</td></tr><tr><td>62dccab9ab715f33761a5315746ed02e48eed2a0</td><td>A Short Note about Kinetics-600 +<br/>Jo˜ao Carreira +</td></tr><tr><td>62d1a31b8acd2141d3a994f2d2ec7a3baf0e6dc4</td><td>Ding et al. EURASIP Journal on Image and Video Processing (2017) 2017:43 +<br/>DOI 10.1186/s13640-017-0188-z +<br/>EURASIP Journal on Image +<br/>and Video Processing +<br/>R ES EAR CH +<br/>Noise-resistant network: a deep-learning +<br/>method for face recognition under noise +<br/>Open Access +</td></tr><tr><td>62694828c716af44c300f9ec0c3236e98770d7cf</td><td>Padrón-Rivera, G., Rebolledo-Mendez, G., Parra, P. P., & Huerta-Pacheco, N. S. (2016). Identification of Action Units Related to +<br/>Identification of Action Units Related to Affective States in a Tutoring System +<br/>1Facultad de Estadística e Informática, Universidad Veracruzana, Mexico // 2Universidad Juárez Autónoma de +<br/>for Mathematics +<br/>Huerta-Pacheco1 +<br/>*Corresponding author +</td></tr><tr><td>620339aef06aed07a78f9ed1a057a25433faa58b</td><td></td></tr><tr><td>62b3598b401c807288a113796f424612cc5833ca</td><td></td></tr><tr><td>628a3f027b7646f398c68a680add48c7969ab1d9</td><td>Plan for Final Year Project: +<br/>HKU-Face: A Large Scale Dataset for Deep Face +<br/>Recognition +<br/>3035140108 +<br/>3035141841 +<br/>Introduction +<br/>Face recognition has been one of the most successful techniques in the field of artificial intelligence +<br/>because of its surpassing human-level performance in academic experiments and broad application in +<br/>the industrial world. Gaussian-face[1] and Facenet[2] hold state-of-the-art record using statistical +<br/>method and deep-learning method respectively. What’s more, face recognition has been applied +<br/>in various areas like authority checking and recording, fostering a large number of start-ups like +<br/>Face++. +<br/>Our final year project will deal with the face recognition task by building a large-scaled and carefully- +<br/>filtered dataset. Our project plan specifies our roadmap and current research process. This plan first +<br/>illustrates the significance and potential enhancement in constructing large-scale face dataset for +<br/>both academics and companies. Then objectives to accomplish and related literature review will be +<br/>expressed in detail. Next, methodologies used, scope of our project and challenges faced by us are +<br/>described. The detailed timeline for this project follows as well as a small summary. +<br/>2 Motivation +<br/>Nowadays most of the face recognition tasks are supervised learning tasks which use dataset annotated +<br/>by human beings. This contains mainly two drawbacks: (1) limited size of dataset due to limited +<br/>human effort; (2) accuracy problem resulted from human perceptual bias. +<br/>Parkhi et al.[3] discuss the first problem, showing that giant companies hold private face databases +<br/>with larger size of data (See the comparison in Table 1). Other research institution could only get +<br/>access to public but smaller databases like LFW[4, 5], which acts like a barricade to even higher +<br/>performance. +<br/>Dataset +<br/>IJB-A [6] +<br/>LFW [4, 5] +<br/>YFD [7] +<br/>CelebFaces [8] +<br/>CASIA-WebFace [9] +<br/>MS-Celeb-1M [10] +<br/>Facebook +<br/>Google +<br/>Availability +<br/>public +<br/>public +<br/>public +<br/>public +<br/>public +<br/>public +<br/>private +<br/>private +<br/>identities +<br/>500 +<br/>5K +<br/>1595 +<br/>10K +<br/>10K +<br/>100K +<br/>4K +<br/>8M +<br/>images +<br/>5712 +<br/>13K +<br/>3425 videos +<br/>202K +<br/>500K +<br/>about 10M +<br/>4400K +<br/>100-200M +<br/>Table 1: Face recognition datasets +</td></tr><tr><td>6257a622ed6bd1b8759ae837b50580657e676192</td><td></td></tr><tr><td>626859fe8cafd25da13b19d44d8d9eb6f0918647</td><td>Activity Recognition based on a +<br/>Magnitude-Orientation Stream Network +<br/>Smart Surveillance Interest Group, Department of Computer Science +<br/>Universidade Federal de Minas Gerais, Belo Horizonte, Brazil +</td></tr><tr><td>620e1dbf88069408b008347cd563e16aeeebeb83</td><td></td></tr><tr><td>62007c30f148334fb4d8975f80afe76e5aef8c7f</td><td>Eye In-Painting with Exemplar Generative Adversarial Networks +<br/>Facebook Inc. +<br/>1 Hacker Way, Menlo Park (CA), USA +</td></tr><tr><td>62a30f1b149843860938de6dd6d1874954de24b7</td><td>418 +<br/>Fast Algorithm for Updating the Discriminant Vectors +<br/>of Dual-Space LDA +</td></tr><tr><td>62e0380a86e92709fe2c64e6a71ed94d152c6643</td><td>Facial Emotion Recognition With Expression Energy +<br/>Albert Cruz +<br/>Center for Research in +<br/>Intelligent Systems +<br/>216 Winston Chung Hall +<br/>Center for Research in +<br/>Intelligent Systems +<br/>216 Winston Chung Hall +<br/>Center for Research in +<br/>Intelligent Systems +<br/>216 Winston Chung Hall +<br/>Riverside, CA, 92521-0425, +<br/>Riverside, CA, 92521-0425, +<br/>Riverside, CA, 92521-0425, +<br/>USA +<br/>USA +<br/>USA +</td></tr><tr><td>961a5d5750f18e91e28a767b3cb234a77aac8305</td><td>Face Detection without Bells and Whistles +<br/>1 ESAT-PSI/VISICS, iMinds, KU Leuven, Belgium +<br/>2 MPI Informatics, Saarbrücken, Germany +<br/>3 D-ITET/CVL, ETH Zürich, Switzerland +</td></tr><tr><td>9626bcb3fc7c7df2c5a423ae8d0a046b2f69180c</td><td>UPTEC STS 17033 +<br/>Examensarbete 30 hp +<br/>November 2017 +<br/>A deep learning approach for +<br/>action classification in American +<br/>football video sequences +</td></tr><tr><td>9696b172d66e402a2e9d0a8d2b3f204ad8b98cc4</td><td>J Inf Process Syst, Vol.9, No.1, March 2013 +<br/>pISSN 1976-913X +<br/>eISSN 2092-805X +<br/>Region-Based Facial Expression Recognition in +<br/>Still Images +</td></tr><tr><td>964a3196d44f0fefa7de3403849d22bbafa73886</td><td></td></tr><tr><td>9606b1c88b891d433927b1f841dce44b8d3af066</td><td>Principal Component Analysis with Tensor Train +<br/>Subspace +</td></tr><tr><td>96b1000031c53cd4c1c154013bb722ffd87fa7da</td><td>ContextVP: Fully Context-Aware Video +<br/>Prediction +<br/>1 NVIDIA, Santa Clara, CA, USA +<br/>2 ETH Zurich, Zurich, Switzerland +<br/>3 The Swiss AI Lab IDSIA, Manno, Switzerland +<br/>4 NNAISENSE, Lugano, Switzerland +</td></tr><tr><td>968f472477a8afbadb5d92ff1b9c7fdc89f0c009</td><td>Firefly-based Facial Expression Recognition +</td></tr><tr><td>9636c7d3643fc598dacb83d71f199f1d2cc34415</td><td></td></tr><tr><td>3a2fc58222870d8bed62442c00341e8c0a39ec87</td><td>Probabilistic Local Variation +<br/>Segmentation +<br/>Technion - Computer Science Department - M.Sc. Thesis MSC-2014-02 - 2014 </td></tr><tr><td>3abc833f4d689f37cc8a28f47fb42e32deaa4b17</td><td>Noname manuscript No. +<br/>(will be inserted by the editor) +<br/>Large Scale Retrieval and Generation of Image Descriptions +<br/>Received: date / Accepted: date +</td></tr><tr><td>3a60678ad2b862fa7c27b11f04c93c010cc6c430</td><td>JANUARY-MARCH 2012 +<br/>A Multimodal Database for +<br/>Affect Recognition and Implicit Tagging +</td></tr><tr><td>3a0a839012575ba455f2b84c2d043a35133285f9</td><td>444 +<br/>Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing, pages 444–454, +<br/>Edinburgh, Scotland, UK, July 27–31, 2011. c(cid:13)2011 Association for Computational Linguistics +</td></tr><tr><td>3a9681e2e07be7b40b59c32a49a6ff4c40c962a2</td><td>Biometrics & Biostatistics International Journal +<br/>Comparing treatment means: overlapping standard +<br/>errors, overlapping confidence intervals, and tests of +<br/>hypothesis +</td></tr><tr><td>3a846704ef4792dd329a5c7a2cb8b330ab6b8b4e</td><td>in any current or +<br/>future media, +<br/>for all other uses, +<br/>© 2010 IEEE. Personal use of this material is permitted. Permission from IEEE must be +<br/>obtained +<br/>including +<br/>reprinting/republishing this material for advertising or promotional purposes, creating +<br/>new collective works, for resale or redistribution to servers or lists, or reuse of any +<br/>copyrighted component of this work in other works. +<br/>Pre-print of article that appeared at the IEEE Computer Society Workshop on Biometrics +<br/>2010. +<br/>The published article can be accessed from: +<br/>http://ieeexplore.ieee.org/xpl/freeabs_all.jsp?arnumber=5544597 +</td></tr><tr><td>3a95eea0543cf05670e9ae28092a114e3dc3ab5c</td><td>Constructing the L2-Graph for Robust Subspace +<br/>Learning and Subspace Clustering +</td></tr><tr><td>3a4f522fa9d2c37aeaed232b39fcbe1b64495134</td><td>ISSN (Online) 2321 – 2004 +<br/>ISSN (Print) 2321 – 5526 +<br/> INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH IN ELECTRICAL, ELECTRONICS, INSTRUMENTATION AND CONTROL ENGINEERING +<br/> Vol. 4, Issue 5, May 2016 +<br/>IJIREEICE +<br/>Face Recognition and Retrieval Using Cross +<br/>Age Reference Coding +<br/>Sricharan H S1, Srinidhi K S1, Rajath D N1, Tejas J N1, Chandrakala B M2 +<br/> BE, DSCE, Bangalore1 +<br/>Assistant Professor, DSCE, Bangalore2 +</td></tr><tr><td>54969bcd728b0f2d3285866c86ef0b4797c2a74d</td><td>IEEE TRANSACTION SUBMISSION +<br/>Learning for Video Compression +</td></tr><tr><td>5456166e3bfe78a353df988897ec0bd66cee937f</td><td>Improved Boosting Performance by Exclusion +<br/>of Ambiguous Positive Examples +<br/>Computer Vision and Active Perception, KTH, Stockholm 10800, Sweden +<br/>Keywords: +<br/>Boosting, Image Classification, Algorithm Evaluation, Dataset Pruning, VOC2007. +</td></tr><tr><td>54aacc196ffe49b3450059fccdf7cd3bb6f6f3c3</td><td>A Joint Learning Framework for Attribute Models and Object Descriptions +<br/>Dhruv Mahajan +<br/>Yahoo! Labs, Bangalore, India +</td></tr><tr><td>541bccf19086755f8b5f57fd15177dc49e77d675</td><td></td></tr><tr><td>549c719c4429812dff4d02753d2db11dd490b2ae</td><td>YouTube-BoundingBoxes: A Large High-Precision +<br/>Human-Annotated Data Set for Object Detection in Video +<br/>Google Brain +<br/>Google Brain +<br/>Google Research +<br/>Google Brain +<br/>Google Brain +</td></tr><tr><td>98b2f21db344b8b9f7747feaf86f92558595990c</td><td></td></tr><tr><td>988d1295ec32ce41d06e7cf928f14a3ee079a11e</td><td>Semantic Deep Learning +<br/>September 29, 2015 +</td></tr><tr><td>981449cdd5b820268c0876477419cba50d5d1316</td><td>Learning Deep Features for One-Class +<br/>Classification +</td></tr><tr><td>98127346920bdce9773aba6a2ffc8590b9558a4a</td><td>Noname manuscript No. +<br/>(will be inserted by the editor) +<br/>Efficient Human Action Recognition using +<br/>Histograms of Motion Gradients and +<br/>VLAD with Descriptor Shape Information +<br/>Received: date / Accepted: date +</td></tr><tr><td>982fed5c11e76dfef766ad9ff081bfa25e62415a</td><td></td></tr><tr><td>98519f3f615e7900578bc064a8fb4e5f429f3689</td><td>Dictionary-based Domain Adaptation Methods +<br/>for the Re-identification of Faces +</td></tr><tr><td>9825aa96f204c335ec23c2b872855ce0c98f9046</td><td>International Journal of Ethics in Engineering & Management Education +<br/>Website: www.ijeee.in (ISSN: 2348-4748, Volume 1, Issue 5, May2014) +<br/>FACE AND FACIAL EXPRESSION +<br/>RECOGNITION IN 3-D USING MASKED +<br/>PROJECTION UNDER OCCLUSION +<br/>Jyoti patil * +<br/>M.Tech (CSE) +<br/>GNDEC Bidar-585401 +<br/>BIDAR, INDIA +<br/> M.Tech (CSE) +<br/> GNDEC Bidar- 585401 +<br/> BIDAR, INDIA +<br/> M.Tech (CSE) +<br/> VKIT, Bangalore- 560040 +<br/>BANGALORE, INDIA +</td></tr><tr><td>5334ac0a6438483890d5eef64f6db93f44aacdf4</td><td></td></tr><tr><td>53dd25350d3b3aaf19beb2104f1e389e3442df61</td><td></td></tr><tr><td>530243b61fa5aea19b454b7dbcac9f463ed0460e</td><td></td></tr><tr><td>539ca9db570b5e43be0576bb250e1ba7a727d640</td><td></td></tr><tr><td>53c8cbc4a3a3752a74f79b74370ed8aeed97db85</td><td></td></tr><tr><td>5366573e96a1dadfcd4fd592f83017e378a0e185</td><td>Böhlen, Chandola and Salunkhe +<br/>Server, server in the cloud. +<br/>Who is the fairest in the crowd? +</td></tr><tr><td>533bfb82c54f261e6a2b7ed7d31a2fd679c56d18</td><td>Technical Report MSU-CSE-14-1 +<br/>Unconstrained Face Recognition: Identifying a +<br/>Person of Interest from a Media Collection +</td></tr><tr><td>530ce1097d0681a0f9d3ce877c5ba31617b1d709</td><td></td></tr><tr><td>3fbd68d1268922ee50c92b28bd23ca6669ff87e5</td><td>598 +<br/>IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 10, NO. 4, APRIL 2001 +<br/>A Shape- and Texture-Based Enhanced Fisher +<br/>Classifier for Face Recognition +</td></tr><tr><td>3f22a4383c55ceaafe7d3cfed1b9ef910559d639</td><td>JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 +<br/>Robust Kronecker Component Analysis +</td></tr><tr><td>3fdcc1e2ebcf236e8bb4a6ce7baf2db817f30001</td><td>A top-down approach for a synthetic +<br/>autobiographical memory system +<br/>1Sheffield Centre for Robotics (SCentRo), Univ. of Sheffield, Sheffield, S10 2TN, UK +<br/>2Dept. of Computer Science, Univ. of Sheffield, Sheffield, S1 4DP, UK +<br/>3 CVAP Lab, KTH, Stockholm, Sweden +</td></tr><tr><td>3f848d6424f3d666a1b6dd405a48a35a797dd147</td><td>GHODRATI et al.: IS 2D INFORMATION ENOUGH FOR VIEWPOINT ESTIMATION? +<br/>Is 2D Information Enough For Viewpoint +<br/>Estimation? +<br/>KU Leuven, ESAT - PSI, iMinds +<br/>Leuven, Belgium +</td></tr><tr><td>3fa738ab3c79eacdbfafa4c9950ef74f115a3d84</td><td>DaMN – Discriminative and Mutually Nearest: +<br/>Exploiting Pairwise Category Proximity +<br/>for Video Action Recognition +<br/>1 Center for Research in Computer Vision at UCF, Orlando, USA +<br/>2 Google Research, Mountain View, USA +<br/>http://crcv.ucf.edu/projects/DaMN/ +</td></tr><tr><td>3fb98e76ffd8ba79e1c22eda4d640da0c037e98a</td><td>Convolutional Neural Networks for Crop Yield Prediction using Satellite Images +<br/>H. Russello +</td></tr><tr><td>3f5cf3771446da44d48f1d5ca2121c52975bb3d3</td><td></td></tr><tr><td>3f14b504c2b37a0e8119fbda0eff52efb2eb2461</td><td>5727 +<br/>Joint Facial Action Unit Detection and Feature +<br/>Fusion: A Multi-Conditional Learning Approach +</td></tr><tr><td>3f9a7d690db82cf5c3940fbb06b827ced59ec01e</td><td>VIP: Finding Important People in Images +<br/>Virginia Tech +<br/>Google Inc. +<br/>Virginia Tech +<br/>Project: https://computing.ece.vt.edu/~mclint/vip/ +<br/>Demo: http://cloudcv.org/vip/ +</td></tr><tr><td>3fd90098551bf88c7509521adf1c0ba9b5dfeb57</td><td>Page 1 of 21 +<br/>*****For Peer Review Only***** +<br/>10 +<br/>11 +<br/>12 +<br/>13 +<br/>14 +<br/>15 +<br/>16 +<br/>17 +<br/>18 +<br/>19 +<br/>20 +<br/>21 +<br/>22 +<br/>23 +<br/>24 +<br/>25 +<br/>26 +<br/>27 +<br/>28 +<br/>29 +<br/>30 +<br/>31 +<br/>32 +<br/>33 +<br/>34 +<br/>35 +<br/>36 +<br/>37 +<br/>38 +<br/>39 +<br/>40 +<br/>41 +<br/>42 +<br/>43 +<br/>44 +<br/>45 +<br/>46 +<br/>47 +<br/>48 +<br/>49 +<br/>50 +<br/>51 +<br/>52 +<br/>53 +<br/>54 +<br/>55 +<br/>56 +<br/>57 +<br/>58 +<br/>59 +<br/>60 +<br/>Attribute-Based Classification for Zero-Shot +<br/>Visual Object Categorization +</td></tr><tr><td>3f63f9aaec8ba1fa801d131e3680900680f14139</td><td>Facial Expression Recognition using Local Binary +<br/>Patterns and Kullback Leibler Divergence +<br/>AnushaVupputuri, SukadevMeher +<br/> +<br/>divergence. +<br/>role +</td></tr><tr><td>3f0e0739677eb53a9d16feafc2d9a881b9677b63</td><td>Efficient Two-Stream Motion and Appearance 3D CNNs for +<br/>Video Classification +<br/>ESAT-KU Leuven +<br/>Ali Pazandeh +<br/>Sharif UTech +<br/>ESAT-KU Leuven, ETH Zurich +</td></tr><tr><td>30870ef75aa57e41f54310283c0057451c8c822b</td><td>Overcoming Catastrophic Forgetting with Hard Attention to the Task +</td></tr><tr><td>303065c44cf847849d04da16b8b1d9a120cef73a</td><td></td></tr><tr><td>3046baea53360a8c5653f09f0a31581da384202e</td><td>Deformable Face Alignment via Local +<br/>Measurements and Global Constraints +</td></tr><tr><td>3028690d00bd95f20842d4aec84dc96de1db6e59</td><td>Leveraging Union of Subspace Structure to Improve Constrained Clustering +</td></tr><tr><td>30c96cc041bafa4f480b7b1eb5c45999701fe066</td><td>1090 +<br/>Discrete Cosine Transform Locality-Sensitive +<br/>Hashes for Face Retrieval +</td></tr><tr><td>306957285fea4ce11a14641c3497d01b46095989</td><td>FACE RECOGNITION UNDER VARYING LIGHTING BASED ON +<br/>DERIVATES OF LOG IMAGE +<br/>2ICT-ISVISION Joint R&D Laboratory for Face Recognition, CAS, Beijing 100080, China +<br/>1Graduate School, CAS, Beijing, 100039, China +</td></tr><tr><td>302c9c105d49c1348b8f1d8cc47bead70e2acf08</td><td>This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TCSVT.2017.2710120, IEEE +<br/>Transactions on Circuits and Systems for Video Technology +<br/>IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY +<br/>Unconstrained Face Recognition Using A Set-to-Set +<br/>Distance Measure +</td></tr><tr><td>304a306d2a55ea41c2355bd9310e332fa76b3cb0</td><td></td></tr><tr><td>5e7e055ef9ba6e8566a400a8b1c6d8f827099553</td><td></td></tr><tr><td>5e28673a930131b1ee50d11f69573c17db8fff3e</td><td>Author manuscript, published in "Workshop on Faces in 'Real-Life' Images: Detection, Alignment, and Recognition, Marseille : France +<br/>(2008)" +</td></tr><tr><td>5e6ba16cddd1797853d8898de52c1f1f44a73279</td><td>Face Identification with Second-Order Pooling +</td></tr><tr><td>5e821cb036010bef259046a96fe26e681f20266e</td><td></td></tr><tr><td>5bfc32d9457f43d2488583167af4f3175fdcdc03</td><td>International Journal of Science and Research (IJSR), India Online ISSN: 2319-7064 +<br/>Local Gray Code Pattern (LGCP): A Robust +<br/>Feature Descriptor for Facial Expression +<br/>Recognition +</td></tr><tr><td>5ba7882700718e996d576b58528f1838e5559225</td><td>This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAFFC.2016.2628787, IEEE +<br/>Transactions on Affective Computing +<br/>IEEE TRANSACTIONS ON AFFECTIVE COMPUTING, VOL. X, NO. X, OCTOBER 2016 +<br/>Predicting Personalized Image Emotion +<br/>Perceptions in Social Networks +</td></tr><tr><td>5bb684dfe64171b77df06ba68997fd1e8daffbe1</td><td></td></tr><tr><td>5bae9822d703c585a61575dced83fa2f4dea1c6d</td><td>MOTChallenge 2015: +<br/>Towards a Benchmark for Multi-Target Tracking +</td></tr><tr><td>5babbad3daac5c26503088782fd5b62067b94fa5</td><td>Are You Sure You Want To Do That? +<br/>Classification with Verification +</td></tr><tr><td>5b9d9f5a59c48bc8dd409a1bd5abf1d642463d65</td><td>Evolving Systems. manuscript No. +<br/>(will be inserted by the editor) +<br/>An evolving spatio-temporal approach for gender and age +<br/>group classification with Spiking Neural Networks +<br/>Received: date / Accepted: date +</td></tr><tr><td>5bf70c1afdf4c16fd88687b4cf15580fd2f26102</td><td>Accepted in Pattern Recognition Letters +<br/>Pattern Recognition Letters +<br/>journal homepage: www.elsevier.com +<br/>Residual Codean Autoencoder for Facial Attribute Analysis +<br/>IIIT-Delhi, New Delhi, India +<br/>Article history: +<br/>Received 29 March 2017 +</td></tr><tr><td>5b2cfee6e81ef36507ebf3c305e84e9e0473575a</td><td></td></tr><tr><td>5be3cc1650c918da1c38690812f74573e66b1d32</td><td>Relative Parts: Distinctive Parts for Learning Relative Attributes +<br/>Center for Visual Information Technology, IIIT Hyderabad, India - 500032 +</td></tr><tr><td>5b0ebb8430a04d9259b321fc3c1cc1090b8e600e</td><td></td></tr><tr><td>3765c26362ad1095dfe6744c6d52494ea106a42c</td><td></td></tr><tr><td>3727ac3d50e31a394b200029b2c350073c1b69e3</td><td></td></tr><tr><td>37f2e03c7cbec9ffc35eac51578e7e8fdfee3d4e</td><td>WACV +<br/>#394 +<br/>000 +<br/>001 +<br/>002 +<br/>003 +<br/>004 +<br/>005 +<br/>006 +<br/>007 +<br/>008 +<br/>009 +<br/>010 +<br/>011 +<br/>012 +<br/>013 +<br/>014 +<br/>015 +<br/>016 +<br/>017 +<br/>018 +<br/>019 +<br/>020 +<br/>021 +<br/>022 +<br/>023 +<br/>024 +<br/>025 +<br/>026 +<br/>027 +<br/>028 +<br/>029 +<br/>030 +<br/>031 +<br/>032 +<br/>033 +<br/>034 +<br/>035 +<br/>036 +<br/>037 +<br/>038 +<br/>039 +<br/>040 +<br/>041 +<br/>042 +<br/>043 +<br/>044 +<br/>045 +<br/>046 +<br/>047 +<br/>048 +<br/>049 +<br/>050 +<br/>051 +<br/>052 +<br/>053 +<br/>WACV 2015 Submission #394. CONFIDENTIAL REVIEW COPY. DO NOT DISTRIBUTE. +<br/>Co-operative Pedestrians Group Tracking in Crowded Scenes using an MST +<br/>Approach +<br/>Anonymous WACV submission +<br/>Paper ID 394 +</td></tr><tr><td>377a1be5113f38297716c4bb951ebef7a93f949a</td><td>Dear Faculty, IGERT Fellows, IGERT Associates and Students, +<br/>You are cordially invited to attend a Seminar presented by Albert Cruz. Please +<br/>plan to attend. +<br/> Albert Cruz +<br/>IGERT Fellow +<br/>Electrical Engineering +<br/> +<br/>Date: Friday, October 11, 2013 +<br/>Location: Bourns A265 +<br/>Time: 11:00am +<br/>Facial emotion recognition with anisotropic +<br/>inhibited gabor energy histograms +</td></tr><tr><td>377c6563f97e76a4dc836a0bd23d7673492b1aae</td><td></td></tr><tr><td>370e0d9b89518a6b317a9f54f18d5398895a7046</td><td>IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY, VOL. X, NO. X, XXXXXXX 20XX +<br/>Cross-pollination of normalisation techniques +<br/>from speaker to face authentication +<br/>using Gaussian mixture models +<br/>and S´ebastien Marcel, Member, IEEE +</td></tr><tr><td>37eb666b7eb225ffdafc6f318639bea7f0ba9a24</td><td>MSU Technical Report (2014): MSU-CSE-14-5 +<br/>Age, Gender and Race Estimation from +<br/>Unconstrained Face Images +</td></tr><tr><td>375435fb0da220a65ac9e82275a880e1b9f0a557</td><td>This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI +<br/>From Pixels to Response Maps: Discriminative Image +<br/>Filtering for Face Alignment in the Wild +</td></tr><tr><td>37b6d6577541ed991435eaf899a2f82fdd72c790</td><td>Vision-based Human Gender Recognition: A Survey +<br/>Universiti Tunku Abdul Rahman, Kuala Lumpur, Malaysia. +</td></tr><tr><td>370b5757a5379b15e30d619e4d3fb9e8e13f3256</td><td>Labeled Faces in the Wild: A Database for Studying +<br/>Face Recognition in Unconstrained Environments +</td></tr><tr><td>08d2f655361335bdd6c1c901642981e650dff5ec</td><td>This is the published version: +<br/> Arandjelovic, Ognjen and Cipolla, R. 2006, Automatic cast listing in feature‐length films with +<br/>Anisotropic Manifold Space, in CVPR 2006 : Proceedings of the Computer Vision and Pattern +<br/>Recognition Conference 2006, IEEE, Piscataway, New Jersey, pp. 1513‐1520. +<br/> +<br/> http://hdl.handle.net/10536/DRO/DU:30058435 +<br/> Reproduced with the kind permission of the copyright owner. +<br/>Copyright : 2006, IEEE +<br/>Available from Deakin Research Online: +</td></tr><tr><td>08ae100805d7406bf56226e9c3c218d3f9774d19</td><td>Gavrilescu and Vizireanu EURASIP Journal on Image and Video Processing (2017) 2017:59 +<br/>DOI 10.1186/s13640-017-0211-4 +<br/>EURASIP Journal on Image +<br/>and Video Processing +<br/>R ES EAR CH +<br/>Predicting the Sixteen Personality Factors +<br/>(16PF) of an individual by analyzing facial +<br/>features +<br/>Open Access +</td></tr><tr><td>08c18b2f57c8e6a3bfe462e599a6e1ce03005876</td><td>A Least-Squares Framework +<br/>for Component Analysis +</td></tr><tr><td>081a431107eb38812b74a8cd036ca5e97235b499</td><td></td></tr><tr><td>0831a511435fd7d21e0cceddb4a532c35700a622</td><td></td></tr><tr><td>080c204edff49bf85b335d3d416c5e734a861151</td><td>CLAD: A Complex and Long Activities +<br/>Dataset with Rich Crowdsourced +<br/>Annotations +<br/>Journal Title +<br/>XX(X):1–6 +<br/>c(cid:13)The Author(s) 2016 +<br/>Reprints and permission: +<br/>sagepub.co.uk/journalsPermissions.nav +<br/>DOI: 10.1177/ToBeAssigned +<br/>www.sagepub.com/ +</td></tr><tr><td>08f4832507259ded9700de81f5fd462caf0d5be8</td><td>International Journal of Computer Applications (0975 – 8887) +<br/>Volume 118 – No.14, May 2015 +<br/>Geometric Approach for Human Emotion +<br/>Recognition using Facial Expression +<br/>S. S. Bavkar +<br/>Assistant Professor +<br/>J. S. Rangole +<br/>Assistant Professor +<br/>V. U. Deshmukh +<br/>Assistant Professor +</td></tr><tr><td>08d40ee6e1c0060d3b706b6b627e03d4b123377a</td><td>Human Action Localization +<br/>with Sparse Spatial Supervision +</td></tr><tr><td>08c1f8f0e69c0e2692a2d51040ef6364fb263a40</td><td></td></tr><tr><td>088aabe3da627432fdccf5077969e3f6402f0a80</td><td>Under review as a conference paper at ICLR 2018 +<br/>CLASSIFIER-TO-GENERATOR ATTACK: ESTIMATION +<br/>OF TRAINING DATA DISTRIBUTION FROM CLASSIFIER +<br/>Anonymous authors +<br/>Paper under double-blind review +</td></tr><tr><td>08903bf161a1e8dec29250a752ce9e2a508a711c</td><td>Joint Dimensionality Reduction and Metric Learning: A Geometric Take +</td></tr><tr><td>08e24f9df3d55364290d626b23f3d42b4772efb6</td><td>ENHANCING FACIAL EXPRESSION CLASSIFICATION BY INFORMATION +<br/>FUSION +<br/>I. Buciu1, Z. Hammal 2, A. Caplier2, N. Nikolaidis 1, and I. Pitas 1 +<br/><b></b><br/>GR-54124, Thessaloniki, Box 451, Greece +<br/>2 Laboratoire des Images et des Signaux / Institut National Polytechnique de Grenoble +<br/>web: http://www.aiia.csd.auth.gr +<br/>38031 Grenoble, France +<br/>web: http://www.lis.inpg.fr +</td></tr><tr><td>0830c9b9f207007d5e07f5269ffba003235e4eff</td><td></td></tr><tr><td>081fb4e97d6bb357506d1b125153111b673cc128</td><td></td></tr><tr><td>0857281a3b6a5faba1405e2c11f4e17191d3824d</td><td>Chude-Olisah et al. EURASIP Journal on Advances in Signal Processing 2014, 2014:102 +<br/>http://asp.eurasipjournals.com/content/2014/1/102 +<br/>R ES EAR CH +<br/>Face recognition via edge-based Gabor feature +<br/>representation for plastic surgery-altered images +<br/>Open Access +</td></tr><tr><td>08f6745bc6c1b0fb68953ea61054bdcdde6d2fc7</td><td>Understanding Kin Relationships in a Photo +</td></tr><tr><td>082ad50ac59fc694ba4369d0f9b87430553b11db</td><td></td></tr><tr><td>6dd052df6b0e89d394192f7f2af4a3e3b8f89875</td><td>International Journal of Engineering and Advanced Technology (IJEAT) +<br/>ISSN: 2249 – 8958, Volume-2, Issue-4, April 2013 +<br/>A literature survey on Facial Expression +<br/>Recognition using Global Features +<br/> +</td></tr><tr><td>6dd5dbb6735846b214be72983e323726ef77c7a9</td><td>Josai Mathematical Monographs +<br/>vol. 7 (2014), pp. 25-40 +<br/>A Survey on Newer Prospective +<br/>Biometric Authentication Modalities +</td></tr><tr><td>6d10beb027fd7213dd4bccf2427e223662e20b7d</td><td></td></tr><tr><td>6dddf1440617bf7acda40d4d75c7fb4bf9517dbb</td><td>JOURNAL OF LATEX CLASS FILES, VOL. XX, NO. X, MM YY +<br/>Beyond Counting: Comparisons of Density Maps for Crowd +<br/>Analysis Tasks - Counting, Detection, and Tracking +</td></tr><tr><td>6de18708218988b0558f6c2f27050bb4659155e4</td><td></td></tr><tr><td>6d91da37627c05150cb40cac323ca12a91965759</td><td></td></tr><tr><td>6d8c9a1759e7204eacb4eeb06567ad0ef4229f93</td><td>Face Alignment Robust to Pose, Expressions and +<br/>Occlusions +</td></tr><tr><td>6d66c98009018ac1512047e6bdfb525c35683b16</td><td>IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 25, NO. 9, SEPTEMBER 2003 +<br/>1063 +<br/>Face Recognition Based on +<br/>Fitting a 3D Morphable Model +</td></tr><tr><td>016cbf0878db5c40566c1fbc237686fbad666a33</td><td></td></tr><tr><td>01bef320b83ac4405b3fc5b1cff788c124109fb9</td><td>de Lausanne +<br/>RLC D1 740, CH-1015 +<br/>Lausanne +<br/>de Lausanne +<br/>RLC D1 740, CH-1015 +<br/>Lausanne +<br/>de Lausanne +<br/>RLC D1 740, CH-1015 +<br/>Lausanne +<br/>Translating Head Motion into Attention - Towards +<br/>Processing of Student’s Body-Language +<br/>CHILI Laboratory +<br/>Łukasz Kidzi´nski +<br/>CHILI Laboratory +<br/>CHILI Laboratory +<br/>École polytechnique fédérale +<br/>École polytechnique fédérale +<br/>École polytechnique fédérale +</td></tr><tr><td>01c8d7a3460422412fba04e7ee14c4f6cdff9ad7</td><td>(IJACSA) International Journal of Advanced Computer Science and Applications, +<br/>Vol. 4, No. 7, 2013 +<br/>Rule Based System for Recognizing Emotions Using +<br/>Multimodal Approach +<br/>Information System +<br/>SBM, SVKM’s NMIMS +<br/>Mumbai, India +<br/> +</td></tr><tr><td>01e12be4097fa8c94cabeef0ad61498c8e7762f2</td><td></td></tr><tr><td>0163d847307fae508d8f40ad193ee542c1e051b4</td><td>JOURNAL OF LATEX CLASS FILES, VOL. 6, NO. 1, JANUARY 2007 +<br/>Classemes and Other Classifier-based +<br/>Features for Efficient Object Categorization +<br/>- Supplementary material - +<br/>1 LOW-LEVEL FEATURES +<br/>We extract the SIFT [1] features for our descriptor +<br/>according to the following pipeline. We first convert +<br/>each image to gray-scale, then we normalize the con- +<br/>trast by forcing the 0.01% of lightest and darkest pixels +<br/>to be mapped to white and black respectively, and +<br/>linearly rescaling the values in between. All images +<br/>exceeding 786,432 pixels of resolution are downsized +<br/>to this maximum value while keeping the aspect ratio. +<br/>The 128-dimensional SIFT descriptors are computed +<br/>from the interest points returned by a DoG detec- +<br/>tor [2]. We finally compute a Bag-Of-Word histogram +<br/>of these descriptors, using a K-means vocabulary of +<br/>500 words. +<br/>2 CLASSEMES +<br/>The LSCOM categories were developed specifically +<br/>for multimedia annotation and retrieval, and have +<br/>been used in the TRECVID video retrieval series. +<br/>We took the LSCOM CYC ontology dated 2006-06-30, +<br/>which contains 2832 unique categories. We removed +</td></tr><tr><td>01c4cf9c7c08f0ad3f386d88725da564f3c54679</td><td>Interpretability Beyond Feature Attribution: +<br/>Quantitative Testing with Concept Activation Vectors (TCAV) +</td></tr><tr><td>017ce398e1eb9f2eed82d0b22fb1c21d3bcf9637</td><td>FACE RECOGNITION WITH HARMONIC DE-LIGHTING +<br/>2ICT-ISVISION Joint R&D Laboratory for Face Recognition, CAS, Beijing, China, 100080 +<br/>1Graduate School, CAS, Beijing, China, 100080 +<br/>Emails: {lyqing, sgshan, wgao}jdl.ac.cn +</td></tr><tr><td>014e3d0fa5248e6f4634dc237e2398160294edce</td><td>Int J Comput Vis manuscript No. +<br/>(will be inserted by the editor) +<br/>What does 2D geometric information really tell us about +<br/>3D face shape? +<br/>Received: date / Accepted: date +</td></tr><tr><td>01beab8f8293a30cf48f52caea6ca0fb721c8489</td><td></td></tr><tr><td>0178929595f505ef7655272cc2c339d7ed0b9507</td><td></td></tr><tr><td>01b4b32c5ef945426b0396d32d2a12c69c282e29</td><td></td></tr><tr><td>0113b302a49de15a1d41ca4750191979ad756d2f</td><td>1424403677/06/$20.00 ©2006 IEEE +<br/>537 +<br/>ICME 2006 +</td></tr><tr><td>064b797aa1da2000640e437cacb97256444dee82</td><td>Coarse-to-fine Face Alignment with Multi-Scale Local Patch Regression +<br/>Megvii Inc. +<br/>Megvii Inc. +<br/>Megvii Inc. +</td></tr><tr><td>06f146dfcde10915d6284981b6b84b85da75acd4</td><td>Scalable Face Image Retrieval using +<br/>Attribute-Enhanced Sparse Codewords +</td></tr><tr><td>0697bd81844d54064d992d3229162fe8afcd82cb</td><td>User-driven mobile robot storyboarding: Learning image interest and +<br/>saliency from pairwise image comparisons +</td></tr><tr><td>06e7e99c1fdb1da60bc3ec0e2a5563d05b63fe32</td><td>WhittleSearch: Image Search with Relative Attribute Feedback +<br/>(Supplementary Material) +<br/>1 Comparative Qualitative Search Results +<br/>We present three qualitative search results for human-generated feedback, in addition to those +<br/>shown in the paper. Each example shows one search iteration, where the 20 reference images are +<br/>randomly selected (rather than ones that match a keyword search, as the image examples in the +<br/>main paper illustrate). For each result, the first figure shows our method and the second figure +<br/>shows the binary feedback result for the corresponding target image. Note that for our method, +<br/>“more/less X” (where X is an attribute) means that the target image is more/less X than the +<br/>reference image which is shown. +<br/>Figures 1 and 2 show results for human-generated relative attribute and binary feedback, re- +<br/>spectively, when both methods are used to target the same “mental image” of a shoe shown in the +<br/>top left bubble. The top right grid of 20 images are the reference images displayed to the user, and +<br/>those outlined and annotated with constraints are the ones chosen by the user to give feedback. +<br/>The bottom row of images in either figure shows the top-ranked images after integrating the user’s +<br/>feedback into the scoring function, revealing the two methods’ respective performance. We see that +<br/>while both methods retrieve high-heeled shoes, only our method retrieves images that are as “open” +<br/>as the target image. This is because using the proposed approach, the user was able to comment +<br/>explicitly on the desired openness property. +</td></tr><tr><td>066d71fcd997033dce4ca58df924397dfe0b5fd1</td><td>(cid:1)(cid:2)(cid:3)(cid:4)(cid:5)(cid:3)(cid:4)(cid:6)(cid:7)(cid:3)(cid:8)(cid:9)(cid:6)(cid:10)(cid:3)(cid:11)(cid:3)(cid:12)(cid:3)(cid:13)(cid:9) +<br/>(cid:3)(cid:4)(cid:14)(cid:6)(cid:15)(cid:16)(cid:3)(cid:17)(cid:18)(cid:3)(cid:11)(cid:5)(cid:19)(cid:4) (cid:20)(cid:5)(cid:11)(cid:21)(cid:6)(cid:3)(cid:6)(cid:22)(cid:9)(cid:20)(cid:6)(cid:10)(cid:9)(cid:11)(cid:9)(cid:8)(cid:11)(cid:5)(cid:19)(cid:4)(cid:6)(cid:23)(cid:17)(cid:24)(cid:19)(cid:2)(cid:5)(cid:11)(cid:21)(cid:25) +<br/>(cid:26)(cid:11)(cid:5)(cid:8)(cid:17)(cid:6)(cid:27)(cid:1)(cid:9)(cid:22)(cid:8)(cid:18)(cid:1)(cid:28)(cid:12)(cid:6)(cid:29)(cid:4)(cid:20)(cid:11)(cid:6)(cid:24)(cid:30)(cid:1)(cid:15)(cid:25)(cid:1)(cid:31)(cid:8)(cid:20)(cid:8) (cid:14)(cid:1)!(cid:8) (cid:8)(cid:6)(cid:4)(cid:1)"(cid:16)(cid:8)(cid:16)(cid:20)(cid:14)(cid:1)(cid:3)(cid:15)(cid:8)(cid:22)(cid:4)(cid:12)(cid:1)(cid:23)(cid:5)(cid:29)(cid:18)(cid:14)(cid:1)(cid:31)(cid:8)(cid:20)(cid:8) (cid:14)(cid:1)(cid:26)!(cid:9)(cid:13)(cid:14)(cid:1)#(cid:17)(cid:8)(cid:6)(cid:5)$(cid:1)(cid:17)(cid:4)(cid:5)%(cid:8)(cid:10)(cid:8)(cid:11)(cid:6)(cid:8)(cid:12)&(cid:30)(cid:8)(cid:16)(cid:15)(cid:15)(cid:21)(cid:27)(cid:15)(cid:17) +<br/>(cid:3)(cid:4)(cid:5)(cid:6)(cid:7)(cid:8)(cid:1)(cid:9)(cid:10)(cid:10)(cid:8)(cid:11)(cid:6)(cid:8)(cid:12)(cid:1)(cid:13)(cid:6)(cid:7)(cid:14) (cid:3)(cid:15)(cid:16)(cid:8)(cid:17)(cid:17)(cid:8)(cid:18)(cid:1)(cid:3)(cid:8)(cid:16)(cid:18)(cid:6)(cid:1)(cid:19)(cid:4)(cid:16)(cid:11)(cid:16)(cid:6)(cid:10)(cid:6)(cid:14)(cid:1)(cid:19)(cid:20)(cid:21)(cid:1)(cid:9)(cid:22)(cid:8)(cid:17)(cid:1)(cid:23)(cid:8)(cid:11)(cid:24)(cid:8)(cid:12)(cid:25)(cid:8)(cid:20)(cid:18) +<br/>(cid:23)(cid:12)(cid:13)(cid:11)(cid:2)(cid:3)(cid:8)(cid:11)$(cid:1)’(cid:16)(cid:6)(cid:11) ((cid:8)((cid:4)(cid:20)(cid:1)(cid:6)(cid:12)(cid:24)(cid:20)(cid:15)(cid:18))(cid:27)(cid:4)(cid:11)(cid:1)(cid:8)(cid:1)(cid:18)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:1)(cid:15)(cid:25)(cid:1)(cid:15)(cid:29)(cid:4)(cid:20)(cid:1)*(cid:14)+,,(cid:1)(cid:27)(cid:15)(cid:5)(cid:15)(cid:20)(cid:1)(cid:6)(cid:17)(cid:8)-(cid:4)(cid:11)(cid:1).(cid:4)(cid:1)(cid:27)(cid:15)(cid:5)(cid:5)(cid:4)(cid:27)(cid:24)(cid:4)(cid:18)(cid:1)(cid:25)(cid:20)(cid:15)(cid:17)(cid:1)+(cid:2)+(cid:1)(cid:18)(cid:6)(cid:25)(cid:25)(cid:4)(cid:20)(cid:4)(cid:12)(cid:24)(cid:1)(cid:16))(cid:17)(cid:8)(cid:12)(cid:1)(cid:25)(cid:8)(cid:27)(cid:4)(cid:11) (cid:6)(cid:12)(cid:1)(cid:8)-(cid:4)(cid:11)(cid:1)(cid:10)(cid:4)(cid:24).(cid:4)(cid:4)(cid:12)(cid:1)/ +<br/>(cid:8)(cid:12)(cid:18) 01(cid:21)(cid:1)2(cid:4)(cid:1)(cid:12)(cid:8)(cid:17)(cid:4)(cid:18)(cid:1)(cid:24)(cid:16)(cid:6)(cid:11)(cid:1)(cid:18)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:1)(cid:24)(cid:16)(cid:4)(cid:1)(cid:26)(cid:20)(cid:8)(cid:12)(cid:6)(cid:8)(cid:12)(cid:1)3(cid:8)(cid:27)(cid:4)(cid:1)(cid:19)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:1)4(cid:26)3(cid:19)(cid:23)5(cid:21)(cid:1)’(cid:15)(cid:1)(cid:4)(cid:29)(cid:8)(cid:5))(cid:8)(cid:24)(cid:4)(cid:1)(cid:24)(cid:16)(cid:4)(cid:1)(cid:18)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:1)(cid:24)(cid:16)(cid:4)(cid:1)(cid:4)6((cid:4)(cid:20)(cid:6)(cid:17)(cid:4)(cid:12)(cid:24)(cid:8)(cid:5)(cid:1)(cid:20)(cid:4)(cid:11))(cid:5)(cid:24)(cid:1)(cid:15)(cid:25)(cid:1)(cid:8)(cid:1)(cid:12)(cid:4).(cid:1)(cid:25)(cid:8)(cid:27)(cid:6)(cid:8)(cid:5)(cid:1) +<br/>(cid:25)(cid:4)(cid:8)(cid:24))(cid:20)(cid:4)(cid:1)(cid:18)(cid:4)(cid:24)(cid:4)(cid:27)(cid:24)(cid:6)(cid:15)(cid:12)(cid:1)(cid:8)(cid:5)-(cid:15)(cid:20)(cid:6)(cid:24)(cid:16)(cid:17)(cid:1)(cid:6)(cid:11)(cid:1)(cid:20)(cid:4)((cid:15)(cid:20)(cid:24)(cid:4)(cid:18)(cid:21) +<br/>(cid:26)(cid:9)(cid:27) (cid:28)(cid:19)(cid:2)(cid:14)(cid:13)$(cid:1)3(cid:8)(cid:27)(cid:4)(cid:1)(cid:26)(cid:17)(cid:8)-(cid:4)(cid:1)(cid:19)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:14)(cid:1)3(cid:8)(cid:27)(cid:6)(cid:8)(cid:5)(cid:1)3(cid:4)(cid:8)(cid:24))(cid:20)(cid:4)(cid:1)(cid:19)(cid:4)(cid:24)(cid:4)(cid:27)(cid:24)(cid:6)(cid:15)(cid:12)(cid:1)(cid:9)(cid:5)-(cid:15)(cid:20)(cid:6)(cid:24)(cid:16)(cid:17)(cid:11)(cid:14)(cid:1)(cid:9)-(cid:4)(cid:1)7(cid:5)(cid:8)(cid:11)(cid:11)(cid:6)(cid:25)(cid:6)(cid:27)(cid:8)(cid:24)(cid:6)(cid:15)(cid:12)(cid:21) +<br/>(cid:29) (cid:1)(cid:4)(cid:11)(cid:2)(cid:19)(cid:14)(cid:18)(cid:8)(cid:11)(cid:5)(cid:19)(cid:4) +<br/>8)(cid:17)(cid:8)(cid:12)(cid:1) (cid:25)(cid:8)(cid:27)(cid:4)(cid:1) (cid:6)(cid:11)(cid:1) (cid:24)(cid:16)(cid:4)(cid:1) (cid:17)(cid:15)(cid:11)(cid:24)(cid:1) (cid:27)(cid:15)(cid:17)(cid:17)(cid:15)(cid:12)(cid:1) (cid:8)(cid:12)(cid:18)(cid:1) )(cid:11)(cid:4)(cid:25))(cid:5)(cid:1) (cid:7)(cid:4)(cid:30)(cid:1) (cid:24)(cid:15)(cid:1) (cid:8)(cid:1) +<br/>((cid:4)(cid:20)(cid:11)(cid:15)(cid:12)9(cid:11)(cid:1) (cid:6)(cid:18)(cid:4)(cid:12)(cid:24)(cid:6)(cid:24)(cid:30)(cid:21)(cid:1) (cid:9)(cid:11)(cid:1) (cid:16))(cid:17)(cid:8)(cid:12)(cid:11)(cid:14)(cid:1) .(cid:4)(cid:1) (cid:8)(cid:20)(cid:4)(cid:1) (cid:8)(cid:10)(cid:5)(cid:4)(cid:1) (cid:24)(cid:15)(cid:1) (cid:27)(cid:8)(cid:24)(cid:4)-(cid:15)(cid:20)(cid:6)(cid:22)(cid:4)(cid:1) (cid:8)(cid:1) +<br/>((cid:4)(cid:20)(cid:11)(cid:15)(cid:12):(cid:11)(cid:1)(cid:8)-(cid:4)(cid:1)-(cid:20)(cid:15))((cid:1)(cid:25)(cid:20)(cid:15)(cid:17)(cid:1)(cid:8)(cid:1)((cid:4)(cid:20)(cid:11)(cid:15)(cid:12):(cid:11)(cid:1)(cid:25)(cid:8)(cid:27)(cid:4)(cid:1)(cid:6)(cid:17)(cid:8)-(cid:4)(cid:1)(cid:8)(cid:12)(cid:18)(cid:1)(cid:8)(cid:20)(cid:4)(cid:1)(cid:15)(cid:25)(cid:24)(cid:4)(cid:12)(cid:1) +<br/>(cid:8)(cid:10)(cid:5)(cid:4)(cid:1)(cid:24)(cid:15)(cid:1)(cid:10)(cid:4)(cid:1);)(cid:6)(cid:24)(cid:4)(cid:1)((cid:20)(cid:4)(cid:27)(cid:6)(cid:11)(cid:4)(cid:1)(cid:6)(cid:12)(cid:1)(cid:24)(cid:16)(cid:6)(cid:11)(cid:1)(cid:4)(cid:11)(cid:24)(cid:6)(cid:17)(cid:8)(cid:24)(cid:6)(cid:15)(cid:12)(cid:1)<(cid:2)=(cid:21)(cid:1)(cid:26)(cid:12)(cid:1)(cid:20)(cid:4)(cid:27)(cid:4)(cid:12)(cid:24)(cid:1)(cid:30)(cid:4)(cid:8)(cid:20)(cid:11)(cid:14)(cid:1) +<br/>(cid:25)(cid:8)(cid:27)(cid:4)(cid:1) (cid:20)(cid:4)(cid:27)(cid:15)-(cid:12)(cid:6)(cid:24)(cid:6)(cid:15)(cid:12)(cid:1) (cid:8)(cid:12)(cid:18)(cid:1) (cid:20)(cid:4)(cid:5)(cid:8)(cid:24)(cid:4)(cid:18)(cid:1) .(cid:15)(cid:20)(cid:7)(cid:11)(cid:1) (cid:16)(cid:8)(cid:29)(cid:4)(cid:1) (cid:20)(cid:4)(cid:27)(cid:4)(cid:6)(cid:29)(cid:4)(cid:18)(cid:1) (cid:11))(cid:10)(cid:11)(cid:24)(cid:8)(cid:12)(cid:24)(cid:6)(cid:8)(cid:5)(cid:1) +<br/>(cid:8)(cid:24)(cid:24)(cid:4)(cid:12)(cid:24)(cid:6)(cid:15)(cid:12)(cid:1) (cid:25)(cid:20)(cid:15)(cid:17)(cid:1) (cid:20)(cid:4)(cid:11)(cid:4)(cid:8)(cid:20)(cid:27)(cid:16)(cid:4)(cid:20)(cid:11)(cid:1) (cid:6)(cid:12)(cid:1) (cid:10)(cid:6)(cid:15)(cid:17)(cid:4)(cid:24)(cid:20)(cid:6)(cid:27)(cid:11)(cid:14)(cid:1) ((cid:8)(cid:24)(cid:24)(cid:4)(cid:20)(cid:12)(cid:1) (cid:20)(cid:4)(cid:27)(cid:15)-(cid:12)(cid:6)(cid:24)(cid:6)(cid:15)(cid:12)(cid:14)(cid:1) +<br/>(cid:8)(cid:12)(cid:18)(cid:1) (cid:27)(cid:15)(cid:17)()(cid:24)(cid:4)(cid:20) (cid:29)(cid:6)(cid:11)(cid:6)(cid:15)(cid:12)(cid:1) (cid:27)(cid:15)(cid:17)(cid:17))(cid:12)(cid:6)(cid:24)(cid:6)(cid:4)(cid:11)(cid:1) </(cid:14)(cid:1) *(cid:14)(cid:1) > (cid:8)(cid:12)(cid:18) 1=(cid:21)(cid:1) ’(cid:16)(cid:4)(cid:11)(cid:4)(cid:1) +<br/>(cid:27)(cid:15)(cid:17)(cid:17)(cid:15)(cid:12)(cid:1)(cid:6)(cid:12)(cid:24)(cid:4)(cid:20)(cid:4)(cid:11)(cid:24)(cid:11)(cid:1)(cid:8)(cid:17)(cid:15)(cid:12)-(cid:1)(cid:20)(cid:4)(cid:11)(cid:4)(cid:8)(cid:20)(cid:27)(cid:16)(cid:4)(cid:20)(cid:11)(cid:1)(cid:17)(cid:15)(cid:24)(cid:6)(cid:29)(cid:8)(cid:24)(cid:4)(cid:18)(cid:1))(cid:11)(cid:1)(cid:24)(cid:15)(cid:1)(cid:27)(cid:15)(cid:5)(cid:5)(cid:4)(cid:27)(cid:24)(cid:1)(cid:8)(cid:1) +<br/>(cid:18)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:1) (cid:15)(cid:25)(cid:1) (cid:25)(cid:8)(cid:27)(cid:6)(cid:8)(cid:5)(cid:1) (cid:6)(cid:17)(cid:8)-(cid:4)(cid:11)(cid:1) (cid:25)(cid:20)(cid:15)(cid:17)(cid:1) ((cid:4)(cid:15)((cid:5)(cid:4)(cid:1) (cid:6)(cid:12)(cid:1) (cid:18)(cid:6)(cid:25)(cid:25)(cid:4)(cid:20)(cid:4)(cid:12)(cid:24)(cid:1) (cid:8)-(cid:4)(cid:11)(cid:21) ’(cid:16)(cid:4)(cid:1) +<br/>(cid:18)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:1)(cid:6)(cid:11)(cid:1)(cid:6)(cid:12)(cid:24)(cid:4)(cid:12)(cid:18)(cid:4)(cid:18)(cid:1)(cid:25)(cid:15)(cid:20)(cid:1)(cid:18)(cid:6)(cid:11)(cid:24)(cid:20)(cid:6)(cid:10))(cid:24)(cid:6)(cid:15)(cid:12)(cid:1)(cid:24)(cid:15)(cid:1)(cid:20)(cid:4)(cid:11)(cid:4)(cid:8)(cid:20)(cid:27)(cid:16)(cid:4)(cid:20)(cid:11)(cid:21) +<br/>’(cid:16)(cid:4)(cid:20)(cid:4)(cid:1) (cid:8)(cid:20)(cid:4)(cid:1) (cid:17)(cid:8)(cid:12)(cid:30)(cid:1) ()(cid:10)(cid:5)(cid:6)(cid:27)(cid:8)(cid:5)(cid:5)(cid:30)(cid:1) (cid:8)(cid:29)(cid:8)(cid:6)(cid:5)(cid:8)(cid:10)(cid:5)(cid:4)(cid:1) (cid:18)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:11)(cid:1) (cid:25)(cid:15)(cid:20)(cid:1) (cid:25)(cid:8)(cid:27)(cid:4)(cid:1) +<br/>(cid:20)(cid:4)(cid:27)(cid:15)-(cid:12)(cid:6)(cid:24)(cid:6)(cid:15)(cid:12)(cid:14)(cid:1) (cid:8)(cid:12)(cid:18)(cid:1) (cid:25)(cid:8)(cid:27)(cid:6)(cid:8)(cid:5)(cid:1) (cid:4)6((cid:20)(cid:4)(cid:11)(cid:11)(cid:6)(cid:15)(cid:12)(cid:1) (cid:8)(cid:12)(cid:8)(cid:5)(cid:30)(cid:11)(cid:6)(cid:11)(cid:21)(cid:1) (cid:23)(cid:4)(cid:11)(cid:6)(cid:18)(cid:4)(cid:1) (cid:8)(cid:10)(cid:15)(cid:29)(cid:4)(cid:1) +<br/>(cid:8)(((cid:5)(cid:6)(cid:27)(cid:8)(cid:24)(cid:6)(cid:15)(cid:12)(cid:11)(cid:14)(cid:1)(cid:26)(cid:20)(cid:8)(cid:12)(cid:6)(cid:8)(cid:12)(cid:1)3(cid:8)(cid:27)(cid:4)(cid:1)(cid:19)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:1)4(cid:26)3(cid:19)(cid:23)5(cid:1)(cid:27)(cid:8)(cid:12)(cid:1)(cid:10)(cid:4)(cid:1))(cid:11)(cid:4)(cid:18)(cid:1)(cid:25)(cid:15)(cid:20)(cid:1)(cid:8)-(cid:4)(cid:1) +<br/>(cid:27)(cid:5)(cid:8)(cid:11)(cid:11)(cid:6)(cid:25)(cid:6)(cid:27)(cid:8)(cid:24)(cid:6)(cid:15)(cid:12)(cid:14)(cid:1) (cid:25)(cid:8)(cid:27)(cid:6)(cid:8)(cid:5)(cid:1) (cid:11))(cid:20)-(cid:4)(cid:20)(cid:30)(cid:14)(cid:1) (cid:20)(cid:8)(cid:27)(cid:4)(cid:1) (cid:18)(cid:4)(cid:24)(cid:4)(cid:27)(cid:24)(cid:6)(cid:15)(cid:12)(cid:1) 4(cid:10)(cid:4)(cid:11)(cid:6)(cid:18)(cid:4)(cid:1) (cid:15)(cid:24)(cid:16)(cid:4)(cid:20)(cid:1) +<br/>(cid:18)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:11)5(cid:14)(cid:1) (cid:11)(cid:24))(cid:18)(cid:30)(cid:6)(cid:12)-(cid:1) (cid:6)(cid:12)(cid:25)(cid:5))(cid:4)(cid:12)(cid:27)(cid:4)(cid:1) (cid:15)(cid:25)(cid:1) (cid:27)(cid:8)(cid:20)(cid:4)(cid:4)(cid:20)(cid:1) (cid:8)(cid:12)(cid:18)(cid:1) (cid:7)(cid:6)(cid:12)(cid:18)(cid:1) (cid:15)(cid:25)(cid:1) (cid:11)(cid:7)(cid:6)(cid:12)(cid:1) (cid:15)(cid:12)(cid:1) +<br/>(cid:8)-(cid:6)(cid:12)-(cid:14)(cid:1)(cid:8)(cid:12)(cid:18)(cid:1)(cid:15)(cid:24)(cid:16)(cid:4)(cid:20)(cid:1)(cid:11)(cid:6)(cid:17)(cid:6)(cid:5)(cid:8)(cid:20)(cid:1)(cid:20)(cid:4)(cid:11)(cid:4)(cid:8)(cid:20)(cid:27)(cid:16)(cid:4)(cid:11)(cid:21) +<br/>(cid:26)(cid:12)(cid:1) (cid:24)(cid:16)(cid:4)(cid:1) (cid:20)(cid:4)(cid:17)(cid:8)(cid:6)(cid:12)(cid:6)(cid:12)-(cid:1) ((cid:8)(cid:20)(cid:24)(cid:11) (cid:18)(cid:4)(cid:24)(cid:8)(cid:6)(cid:5)(cid:11)(cid:1) (cid:15)(cid:25)(cid:1) (cid:24)(cid:16)(cid:4)(cid:1) (cid:4)6(cid:6)(cid:11)(cid:24)(cid:6)(cid:12)-(cid:1) (cid:25)(cid:8)(cid:27)(cid:4)(cid:1) (cid:6)(cid:17)(cid:8)-(cid:4)(cid:1) +<br/>(cid:18)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:11) (cid:8)(cid:12)(cid:18) (cid:24)(cid:16)(cid:4)(cid:1)(cid:26)(cid:20)(cid:8)(cid:12)(cid:6)(cid:8)(cid:12)(cid:1)3(cid:8)(cid:27)(cid:4)(cid:1)(cid:26)(cid:17)(cid:8)-(cid:4)(cid:1)(cid:19)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:1)(cid:6)(cid:11)(cid:1)-(cid:6)(cid:29)(cid:4)(cid:12)(cid:21) (cid:9)(cid:5)(cid:11)(cid:15)(cid:1) +<br/>(cid:24)(cid:16)(cid:4)(cid:1) (cid:18)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:1) (cid:6)(cid:11)(cid:1) (cid:4)(cid:29)(cid:8)(cid:5))(cid:8)(cid:24)(cid:4)(cid:18)(cid:1) (cid:10)(cid:30)(cid:1) (cid:8)(((cid:5)(cid:30)(cid:6)(cid:12)- (cid:8)(cid:1) (cid:12)(cid:4).(cid:1) (cid:25)(cid:8)(cid:27)(cid:6)(cid:8)(cid:5)(cid:1) (cid:25)(cid:4)(cid:8)(cid:24))(cid:20)(cid:4)(cid:1) +<br/>(cid:18)(cid:4)(cid:24)(cid:4)(cid:27)(cid:24)(cid:6)(cid:15)(cid:12)(cid:1)(cid:8)(cid:5)-(cid:15)(cid:20)(cid:6)(cid:24)(cid:16)(cid:17)(cid:21)(cid:1) +<br/>(cid:30) (cid:15)(cid:31)(cid:5)(cid:13)(cid:11)(cid:5)(cid:4)(cid:24)(cid:6)(cid:7)(cid:3)(cid:8)(cid:9)(cid:6)(cid:1)(cid:25)(cid:3)(cid:24)(cid:9)(cid:6)(cid:10)(cid:3)(cid:11)(cid:3)(cid:12)(cid:3)(cid:13)(cid:9)(cid:13) +<br/>(cid:3)(cid:8)(cid:12)(cid:30)(cid:1) (cid:25)(cid:8)(cid:27)(cid:4)(cid:1) (cid:18)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:11)(cid:1) (cid:8)(cid:20)(cid:4)(cid:1) (cid:20)(cid:4)(cid:27)(cid:15)(cid:20)(cid:18)(cid:4)(cid:18)(cid:1) )(cid:12)(cid:18)(cid:4)(cid:20)(cid:1) (cid:8)(cid:1) (cid:29)(cid:8)(cid:20)(cid:6)(cid:4)(cid:24)(cid:30)(cid:1) (cid:15)(cid:25)(cid:1) +<br/>(cid:27)(cid:15)(cid:12)(cid:18)(cid:6)(cid:24)(cid:6)(cid:15)(cid:12)(cid:11)(cid:1)(cid:8)(cid:12)(cid:18)(cid:1).(cid:6)(cid:24)(cid:16)(cid:1)(cid:29)(cid:8)(cid:20)(cid:6)(cid:15))(cid:11)(cid:1)(cid:8)(((cid:5)(cid:6)(cid:27)(cid:8)(cid:24)(cid:6)(cid:15)(cid:12)(cid:11)(cid:1)(cid:6)(cid:12)(cid:1)(cid:17)(cid:6)(cid:12)(cid:18)(cid:21)(cid:1)(cid:9)(cid:5)(cid:15)(cid:12)-(cid:1).(cid:6)(cid:24)(cid:16)(cid:1) +<br/>(cid:24)(cid:16)(cid:4)(cid:1) (cid:18)(cid:4)(cid:29)(cid:4)(cid:5)(cid:15)((cid:17)(cid:4)(cid:12)(cid:24)(cid:1) (cid:15)(cid:25)(cid:1) (cid:25)(cid:8)(cid:27)(cid:4)(cid:1) (cid:20)(cid:4)(cid:27)(cid:15)-(cid:12)(cid:6)(cid:24)(cid:6)(cid:15)(cid:12)(cid:1) (cid:8)(cid:12)(cid:18)(cid:1) (cid:25)(cid:8)(cid:27)(cid:6)(cid:8)(cid:5)(cid:1) (cid:4)6((cid:20)(cid:4)(cid:11)(cid:11)(cid:6)(cid:15)(cid:12)(cid:1) +<br/>(cid:8)(cid:12)(cid:8)(cid:5)(cid:30)(cid:11)(cid:6)(cid:11)(cid:1) (cid:8)(cid:5)-(cid:15)(cid:20)(cid:6)(cid:24)(cid:16)(cid:17)(cid:11)(cid:14)(cid:1) (cid:8)(cid:1) (cid:27)(cid:15)(cid:17)((cid:8)(cid:20)(cid:8)(cid:24)(cid:6)(cid:29)(cid:4)(cid:5)(cid:30)(cid:1) (cid:5)(cid:8)(cid:20)-(cid:4)(cid:1) (cid:12))(cid:17)(cid:10)(cid:4)(cid:20)(cid:1) (cid:15)(cid:25)(cid:1) (cid:25)(cid:8)(cid:27)(cid:4)(cid:1) +<br/>A(cid:8)(cid:5)(cid:4)(cid:1)<0=(cid:14)(cid:1)(cid:3)(cid:26)’(cid:1)<B=(cid:14)(cid:1)C(cid:9)33#(cid:1)<(cid:2),=(cid:14)(cid:1)(cid:8)(cid:12)(cid:18)(cid:1)(cid:17)(cid:8)(cid:12)(cid:30)(cid:1)(cid:15)(cid:24)(cid:16)(cid:4)(cid:20)(cid:1)(cid:18)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:11)(cid:1)<(cid:2)(cid:2)(cid:14)(cid:1) +<br/>(cid:2)/=(cid:21)(cid:1)8(cid:4)(cid:20)(cid:4)(cid:1)3#!#’(cid:1)<(cid:2)*= (cid:8)(cid:12)(cid:18)(cid:1)3DE(cid:13)#’(cid:1)<(cid:2)>=(cid:1)(cid:8)(cid:20)(cid:4)(cid:1)(cid:20)(cid:4)(cid:29)(cid:6)(cid:4).(cid:4)(cid:18)(cid:21) +<br/>(cid:30) (cid:29) (cid:7)(cid:15)!(cid:15)"(cid:6)(cid:10)(cid:3)(cid:11)(cid:3)(cid:12)(cid:3)(cid:13)(cid:9) +<br/>’(cid:16)(cid:4)(cid:1) 3(cid:8)(cid:27)(cid:6)(cid:8)(cid:5)(cid:1) !(cid:4)(cid:27)(cid:15)-(cid:12)(cid:6)(cid:24)(cid:6)(cid:15)(cid:12)(cid:1) ’(cid:4)(cid:27)(cid:16)(cid:12)(cid:15)(cid:5)(cid:15)-(cid:30)(cid:1) 43#!#’5(cid:1) (cid:18)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:1) +<br/>.(cid:8)(cid:11)(cid:1)(cid:27)(cid:15)(cid:5)(cid:5)(cid:4)(cid:27)(cid:24)(cid:4)(cid:18)(cid:1) (cid:8)(cid:24)(cid:1) D(cid:4)(cid:15)(cid:20)-(cid:4)(cid:1)(cid:3)(cid:8)(cid:11)(cid:15)(cid:12)(cid:1) (cid:28)(cid:12)(cid:6)(cid:29)(cid:4)(cid:20)(cid:11)(cid:6)(cid:24)(cid:30)(cid:1) (cid:8)(cid:12)(cid:18)(cid:1)(cid:24)(cid:16)(cid:4)(cid:1) (cid:28)"(cid:1) (cid:9)(cid:20)(cid:17)(cid:30)(cid:1) +<br/>!(cid:4)(cid:11)(cid:4)(cid:8)(cid:20)(cid:27)(cid:16)(cid:1)F(cid:8)(cid:10)(cid:15)(cid:20)(cid:8)(cid:24)(cid:15)(cid:20)(cid:30)(cid:1)(cid:25)(cid:8)(cid:27)(cid:6)(cid:5)(cid:6)(cid:24)(cid:6)(cid:4)(cid:11)(cid:1) (cid:8)(cid:11)(cid:1)((cid:8)(cid:20)(cid:24)(cid:1)(cid:15)(cid:25)(cid:1) (cid:24)(cid:16)(cid:4)(cid:1)3#!#’(cid:1) ((cid:20)(cid:15)-(cid:20)(cid:8)(cid:17)(cid:1) +<br/><(cid:2)*=(cid:21)(cid:1)(cid:26)(cid:12)(cid:1)3#!#’(cid:1)(cid:18)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:1)(cid:6)(cid:17)(cid:8)-(cid:4)(cid:11)(cid:1)(cid:15)(cid:25)(cid:1)(cid:2)(cid:2)BB(cid:1)(cid:11))(cid:10) (cid:4)(cid:27)(cid:24)(cid:1)(cid:4)6(cid:6)(cid:11)(cid:24)(cid:1)(cid:6)(cid:12)(cid:1)BE/, +<br/>(cid:18)(cid:6)(cid:25)(cid:25)(cid:4)(cid:20)(cid:4)(cid:12)(cid:24)(cid:1) ((cid:15)(cid:11)(cid:4)(cid:11)(cid:14)(cid:1) /(cid:1) +<br/>(cid:25)(cid:8)(cid:27)(cid:6)(cid:8)(cid:5)(cid:1) (cid:4)6((cid:20)(cid:4)(cid:11)(cid:11)(cid:6)(cid:15)(cid:12)(cid:11)(cid:1) (cid:8)(cid:12)(cid:18)(cid:1) /(cid:1) (cid:18)(cid:6)(cid:25)(cid:25)(cid:4)(cid:20)(cid:4)(cid:12)(cid:24)(cid:1) +<br/>(cid:6)(cid:5)(cid:5))(cid:17)(cid:6)(cid:12)(cid:8)(cid:24)(cid:6)(cid:15)(cid:12)(cid:11)(cid:1)(cid:6)(cid:12)(cid:1)/(cid:1)(cid:18)(cid:6)(cid:25)(cid:25)(cid:4)(cid:20)(cid:4)(cid:12)(cid:24)(cid:1)(cid:24)(cid:6)(cid:17)(cid:4)(cid:11)(cid:21)(cid:1)(cid:1)’(cid:16)(cid:4)(cid:20)(cid:4)(cid:1)(cid:8)(cid:20)(cid:4) (cid:2)>(cid:14),1(cid:2)(cid:1)(cid:6)(cid:17)(cid:8)-(cid:4)(cid:11)(cid:1)(cid:6)(cid:12)(cid:1) +<br/>/1+G*0>(cid:1)((cid:6)6(cid:4)(cid:5)(cid:11)(cid:1)(cid:6)(cid:12)(cid:1)(cid:11)(cid:6)(cid:22)(cid:4)(cid:21)(cid:1)(cid:26)(cid:17)(cid:8)-(cid:4)(cid:11)(cid:1).(cid:4)(cid:20)(cid:4)(cid:1)(cid:27)(cid:15)(cid:5)(cid:5)(cid:4)(cid:27)(cid:24)(cid:4)(cid:18)(cid:1)(cid:8)(cid:24)(cid:1)(cid:24)(cid:16)(cid:4)(cid:1)(cid:25)(cid:15)(cid:5)(cid:5)(cid:15).(cid:6)(cid:12)-(cid:1) +<br/>((cid:15)(cid:11)(cid:4)(cid:11)$(cid:1)(cid:20)(cid:6)-(cid:16)(cid:24)(cid:1)(cid:8)(cid:12)(cid:18)(cid:1)(cid:5)(cid:4)(cid:25)(cid:24)(cid:1)((cid:20)(cid:15)(cid:25)(cid:6)(cid:5)(cid:4)(cid:14)(cid:1)(cid:20)(cid:6)-(cid:16)(cid:24)(cid:1)(cid:8)(cid:12)(cid:18)(cid:1)(cid:5)(cid:4)(cid:25)(cid:24)(cid:1);)(cid:8)(cid:20)(cid:24)(cid:4)(cid:20)(cid:1)((cid:20)(cid:15)(cid:25)(cid:6)(cid:5)(cid:4)(cid:1)(cid:8)(cid:12)(cid:18)(cid:1) +<br/>(cid:20)(cid:6)-(cid:16)(cid:24)(cid:1) (cid:8)(cid:12)(cid:18)(cid:1) (cid:5)(cid:4)(cid:25)(cid:24)(cid:1) (cid:16)(cid:8)(cid:5)(cid:25)(cid:1) ((cid:20)(cid:15)(cid:25)(cid:6)(cid:5)(cid:4)(cid:21)(cid:1) (cid:26)(cid:12)(cid:1) (cid:24)(cid:16)(cid:4)(cid:11)(cid:4)(cid:1) (cid:27)(cid:8)(cid:24)(cid:4)-(cid:15)(cid:20)(cid:6)(cid:4)(cid:11)(cid:1) (cid:6)(cid:17)(cid:8)-(cid:4)(cid:11)(cid:1) .(cid:4)(cid:20)(cid:4)(cid:1) +<br/>(cid:20)(cid:4)(cid:27)(cid:15)(cid:20)(cid:18)(cid:4)(cid:18)(cid:1)(cid:25)(cid:15)(cid:20)(cid:1)1,0(cid:1)(cid:24)(cid:15)(cid:1)B0,(cid:1)(cid:11))(cid:10) (cid:4)(cid:27)(cid:24)(cid:11)(cid:21) +<br/>(cid:30) (cid:30)(cid:6)(cid:7)#$(cid:22)(cid:15)"(cid:6)(cid:23)(cid:24)(cid:5)(cid:4)(cid:24)(cid:6)(cid:10)(cid:3)(cid:11)(cid:3)(cid:12)(cid:3)(cid:13)(cid:9) +<br/>(cid:2) +<br/>’(cid:16)(cid:4)(cid:1)3DE(cid:13)#’(cid:1)(cid:9)-(cid:6)(cid:12)-(cid:1)(cid:19)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:1).(cid:8)(cid:11)(cid:1)-(cid:4)(cid:12)(cid:4)(cid:20)(cid:8)(cid:24)(cid:4)(cid:18)(cid:1)(cid:8)(cid:11)(cid:1)((cid:8)(cid:20)(cid:24)(cid:1)(cid:15)(cid:25)(cid:1)(cid:24)(cid:16)(cid:4)(cid:1) +<br/>#)(cid:20)(cid:15)((cid:4)(cid:8)(cid:12)(cid:1) (cid:28)(cid:12)(cid:6)(cid:15)(cid:12)(cid:1) ((cid:20)(cid:15) (cid:4)(cid:27)(cid:24)(cid:1) 3DE(cid:13)#’(cid:1) +<br/>43(cid:8)(cid:27)(cid:4)(cid:1) (cid:8)(cid:12)(cid:18)(cid:1) D(cid:4)(cid:11)(cid:24))(cid:20)(cid:4)(cid:1) +<br/>!(cid:4)(cid:27)(cid:15)-(cid:12)(cid:6)(cid:24)(cid:6)(cid:15)(cid:12)(cid:1) !(cid:4)(cid:11)(cid:4)(cid:8)(cid:20)(cid:27)(cid:16)(cid:1) (cid:13)(cid:4)(cid:24).(cid:15)(cid:20)(cid:7)5(cid:21)’(cid:16)(cid:6)(cid:11)(cid:1) (cid:18)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:1) (cid:6)(cid:11)(cid:1) (cid:27)(cid:15)(cid:12)(cid:24)(cid:8)(cid:6)(cid:12)(cid:6)(cid:12)-(cid:1) +<br/>(cid:2),,/(cid:1) (cid:11)(cid:27)(cid:8)(cid:12)(cid:12)(cid:4)(cid:18)(cid:1) (cid:25)(cid:8)(cid:27)(cid:4)(cid:1) (cid:6)(cid:17)(cid:8)-(cid:4)(cid:11)(cid:1) (cid:11)(cid:16)(cid:15).(cid:6)(cid:12)-(cid:1) 0/(cid:1) (cid:11))(cid:10) (cid:4)(cid:27)(cid:24)(cid:11)(cid:1) (cid:8)(cid:24)(cid:1) (cid:18)(cid:6)(cid:25)(cid:25)(cid:4)(cid:20)(cid:4)(cid:12)(cid:24)(cid:1) +<br/>(cid:8)-(cid:4)(cid:11)(cid:21)(cid:1)(cid:26)(cid:17)(cid:8)-(cid:4)(cid:11)(cid:1)(cid:16)(cid:8)(cid:29)(cid:4)(cid:1)(cid:29)(cid:8)(cid:20)(cid:30)(cid:6)(cid:12)-(cid:1)(cid:20)(cid:4)(cid:11)(cid:15)(cid:5))(cid:24)(cid:6)(cid:15)(cid:12)?(cid:1)(cid:8)(((cid:20)(cid:15)6(cid:6)(cid:17)(cid:8)(cid:24)(cid:4)(cid:5)(cid:30)(cid:1)>,,G1,, +<br/>((cid:6)6(cid:4)(cid:5)(cid:11)(cid:21)(cid:1) ’(cid:16)(cid:4)(cid:1) (cid:18)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:1) .(cid:8)(cid:11)(cid:1) (cid:18)(cid:4)(cid:29)(cid:4)(cid:5)(cid:15)((cid:4)(cid:18)(cid:1) (cid:6)(cid:12)(cid:1) (cid:8)(cid:12)(cid:1) (cid:8)(cid:24)(cid:24)(cid:4)(cid:17)((cid:24)(cid:1) (cid:24)(cid:15)(cid:1) (cid:8)(cid:11)(cid:11)(cid:6)(cid:11)(cid:24)(cid:1) +<br/>(cid:20)(cid:4)(cid:11)(cid:4)(cid:8)(cid:20)(cid:27)(cid:16)(cid:4)(cid:20)(cid:11)(cid:1) .(cid:16)(cid:15)(cid:1) (cid:6)(cid:12)(cid:29)(cid:4)(cid:11)(cid:24)(cid:6)-(cid:8)(cid:24)(cid:4)(cid:1) (cid:24)(cid:16)(cid:4)(cid:1) (cid:4)(cid:25)(cid:25)(cid:4)(cid:27)(cid:24)(cid:11)(cid:1) (cid:15)(cid:25)(cid:1) (cid:8)-(cid:6)(cid:12)-(cid:1) (cid:15)(cid:12)(cid:1) (cid:25)(cid:8)(cid:27)(cid:6)(cid:8)(cid:5)(cid:1) +<br/>(cid:8)(((cid:4)(cid:8)(cid:20)(cid:8)(cid:12)(cid:27)(cid:4)(cid:1)<(cid:2)> =(cid:21) +<br/>(cid:30) % (cid:22)(cid:9)(cid:9)(cid:14)(cid:6)(cid:7)(cid:19)(cid:2)(cid:6)(cid:23)(cid:6)(cid:22)(cid:9)(cid:20)(cid:6)(cid:10)(cid:3)(cid:11)(cid:3)(cid:12)(cid:3)(cid:13)(cid:9) +<br/>(cid:26)(cid:12)(cid:1)(cid:15)(cid:20)(cid:18)(cid:4)(cid:20)(cid:1)(cid:24)(cid:15)(cid:1) (cid:10))(cid:6)(cid:5)(cid:18)(cid:14)(cid:1) (cid:24)(cid:20)(cid:8)(cid:6)(cid:12)(cid:1) (cid:8)(cid:12)(cid:18)(cid:1) (cid:20)(cid:4)(cid:5)(cid:6)(cid:8)(cid:10)(cid:5)(cid:30)(cid:1) (cid:24)(cid:4)(cid:11)(cid:24)(cid:1) (cid:8)-(cid:4)(cid:1) (cid:27)(cid:5)(cid:8)(cid:11)(cid:11)(cid:6)(cid:25)(cid:6)(cid:27)(cid:8)(cid:24)(cid:6)(cid:15)(cid:12)(cid:1) +<br/>(cid:8)(cid:5)-(cid:15)(cid:20)(cid:6)(cid:24)(cid:16)(cid:17)(cid:11)(cid:14)(cid:1)(cid:18)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:11)(cid:1).(cid:6)(cid:24)(cid:16)(cid:1)(cid:27)(cid:15)(cid:12)(cid:24)(cid:20)(cid:15)(cid:5)(cid:5)(cid:4)(cid:18)(cid:1)(cid:29)(cid:8)(cid:20)(cid:6)(cid:8)(cid:24)(cid:6)(cid:15)(cid:12)(cid:11)(cid:1)(cid:15)(cid:25)(cid:1)(cid:25)(cid:8)(cid:27)(cid:24)(cid:15)(cid:20)(cid:11)(cid:1)(cid:11))(cid:27)(cid:16)(cid:1) +<br/>(cid:8)(cid:11)(cid:1)(cid:8)-(cid:4)(cid:14)(cid:1)(cid:25)(cid:8)(cid:27)(cid:4)(cid:1)((cid:15)(cid:11)(cid:4)(cid:14)(cid:1)(cid:25)(cid:8)(cid:27)(cid:6)(cid:8)(cid:5)(cid:1)(cid:4)6((cid:20)(cid:4)(cid:11)(cid:11)(cid:6)(cid:15)(cid:12)(cid:14)(cid:1)(cid:15)(cid:27)(cid:27)(cid:5))(cid:11)(cid:6)(cid:15)(cid:12)(cid:1)(cid:25)(cid:8)(cid:27)(cid:6)(cid:8)(cid:5)(cid:1)(cid:16)(cid:8)(cid:6)(cid:20)(cid:14)(cid:1)(cid:8)(cid:12)(cid:18)(cid:1) +<br/>(cid:6)(cid:5)(cid:5))(cid:17)(cid:6)(cid:12)(cid:8)(cid:24)(cid:6)(cid:15)(cid:12)(cid:1) (cid:6)(cid:11)(cid:1) (cid:12)(cid:4)(cid:4)(cid:18)(cid:4)(cid:18)(cid:21)(cid:1) (cid:26)(cid:12)(cid:1) (cid:11)((cid:6)(cid:24)(cid:4)(cid:1) (cid:15)(cid:25)(cid:1) (cid:29)(cid:8)(cid:20)(cid:6)(cid:15))(cid:11)(cid:1) (cid:18)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:11)(cid:14)(cid:1) (cid:24)(cid:16)(cid:4)(cid:20)(cid:4)(cid:1) (cid:6)(cid:11)(cid:1) +<br/>(cid:12)(cid:15)(cid:24)(cid:1)(cid:8)(cid:12)(cid:1)(cid:8)(((cid:20)(cid:15)((cid:20)(cid:6)(cid:8)(cid:24)(cid:4)(cid:1)(cid:15)(cid:12)(cid:4)(cid:1)(cid:25)(cid:15)(cid:20)(cid:1)(cid:8)-(cid:4)(cid:1)(cid:27)(cid:5)(cid:8)(cid:11)(cid:11)(cid:6)(cid:25)(cid:6)(cid:27)(cid:8)(cid:24)(cid:6)(cid:15)(cid:12)(cid:21)(cid:1)(cid:3)(cid:15)(cid:11)(cid:24)(cid:1)(cid:27))(cid:20)(cid:20)(cid:4)(cid:12)(cid:24)(cid:1)(cid:18)(cid:8)(cid:24)(cid:8)(cid:1) +<br/>(cid:10)(cid:8)(cid:11)(cid:4)(cid:11)(cid:1)(cid:18)(cid:15)(cid:12):(cid:24)(cid:1)(cid:16)(cid:8)(cid:29)(cid:4)(cid:1)(cid:6)(cid:17)(cid:8)-(cid:4)(cid:11)(cid:1)(cid:15)(cid:25)(cid:1)((cid:4)(cid:15)((cid:5)(cid:4)(cid:1)(cid:6)(cid:12)(cid:1)(cid:18)(cid:6)(cid:25)(cid:25)(cid:4)(cid:20)(cid:4)(cid:12)(cid:24)(cid:1)(cid:8)-(cid:4)(cid:11)(cid:14)(cid:1)(cid:8)(cid:12)(cid:18)(cid:1)(cid:6)(cid:25)(cid:1)(cid:24)(cid:16)(cid:4)(cid:30)(cid:1) +<br/>(cid:16)(cid:8)(cid:29)(cid:4)(cid:14)(cid:1) (cid:24)(cid:16)(cid:4)(cid:30)(cid:1) (cid:18)(cid:15)(cid:1) (cid:12)(cid:15)(cid:24)(cid:1) (cid:17)(cid:4)(cid:12)(cid:24)(cid:6)(cid:15)(cid:12)(cid:1) (cid:24)(cid:16)(cid:4)(cid:6)(cid:20)(cid:1) (cid:8)-(cid:4)(cid:11)(cid:21)(cid:1) 3DE(cid:13)#’(cid:1) (cid:18)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:1) +<br/>(cid:27)(cid:15)(cid:12)(cid:24)(cid:8)(cid:6)(cid:12)(cid:11)(cid:1) (cid:11)(cid:27)(cid:8)(cid:12)(cid:12)(cid:4)(cid:18)(cid:1) (cid:6)(cid:17)(cid:8)-(cid:4)(cid:11)(cid:1) (cid:15)(cid:25)(cid:1) ((cid:4)(cid:20)(cid:11)(cid:15)(cid:12)(cid:11)(cid:1) .(cid:6)(cid:24)(cid:16)(cid:1) (cid:17)(cid:4)(cid:12)(cid:24)(cid:6)(cid:15)(cid:12)(cid:6)(cid:12)-(cid:1) (cid:24)(cid:16)(cid:4)(cid:6)(cid:20)(cid:1) +<br/>(cid:8)-(cid:4)(cid:11)?(cid:1)(cid:10))(cid:24)(cid:1)(cid:18)(cid:6)(cid:25)(cid:25)(cid:4)(cid:20)(cid:4)(cid:12)(cid:24)(cid:1)(cid:5)(cid:6)-(cid:16)(cid:24)(cid:6)(cid:12)-(cid:1)(cid:27)(cid:15)(cid:12)(cid:18)(cid:6)(cid:24)(cid:6)(cid:15)(cid:12)(cid:11)(cid:14)(cid:1)(cid:10)(cid:8)(cid:27)(cid:7)-(cid:20)(cid:15))(cid:12)(cid:18)(cid:14)(cid:1)((cid:15)(cid:11)(cid:4)(cid:11)(cid:1)(cid:8)(cid:12)(cid:18)(cid:1) +<br/>(cid:4)6((cid:20)(cid:4)(cid:11)(cid:11)(cid:6)(cid:15)(cid:12)(cid:11)(cid:21)(cid:1)(cid:23)(cid:30)(cid:1)(cid:11)(cid:24))(cid:18)(cid:30)(cid:6)(cid:12)-(cid:1)(cid:15)(cid:24)(cid:16)(cid:4)(cid:20)(cid:1)(cid:18)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:11)(cid:1)(cid:6)(cid:24)(cid:1) .(cid:8)(cid:11)(cid:1) (cid:27)(cid:15)(cid:12)(cid:27)(cid:5))(cid:18)(cid:4)(cid:18)(cid:1)(cid:24)(cid:15)(cid:1) +<br/>((cid:20)(cid:15)(cid:29)(cid:6)(cid:18)(cid:4)(cid:1) (cid:8)(cid:1) (cid:18)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:1) .(cid:6)(cid:24)(cid:16)(cid:1) (cid:27)(cid:15)(cid:12)(cid:18)(cid:6)(cid:24)(cid:6)(cid:15)(cid:12)(cid:11)(cid:1) (cid:15)(cid:25)(cid:1) (cid:8)(cid:12)(cid:1) (cid:8)-(cid:4)(cid:1) (cid:27)(cid:5)(cid:8)(cid:11)(cid:11)(cid:6)(cid:25)(cid:6)(cid:27)(cid:8)(cid:24)(cid:6)(cid:15)(cid:12)(cid:11)(cid:1) +<br/>((cid:20)(cid:15) (cid:4)(cid:27)(cid:24)(cid:21)(cid:1) (cid:9)-(cid:4)(cid:14)(cid:1) (cid:4)(cid:12)(cid:15))-(cid:16)(cid:1) (cid:20)(cid:4)(cid:11)(cid:15)(cid:5))(cid:24)(cid:6)(cid:15)(cid:12)(cid:1) (cid:25)(cid:15)(cid:20)(cid:1) .(cid:20)(cid:6)(cid:12)(cid:7)(cid:5)(cid:4)(cid:1) (cid:8)(cid:12)(cid:8)(cid:5)(cid:30)(cid:11)(cid:6)(cid:11)(cid:1) (cid:8)(cid:12)(cid:18)(cid:1) +<br/>(cid:25)(cid:20)(cid:15)(cid:12)(cid:24)(cid:8)(cid:5)(cid:1)((cid:15)(cid:11)(cid:4)(cid:11)(cid:1)(cid:8)(cid:20)(cid:4)(cid:1)(cid:10)(cid:8)(cid:11)(cid:6)(cid:27)(cid:1)(cid:12)(cid:4)(cid:4)(cid:18)(cid:11)(cid:1)(cid:25)(cid:15)(cid:20)(cid:1)(cid:24)(cid:16)(cid:6)(cid:11)(cid:1)(cid:25)(cid:6)(cid:4)(cid:5)(cid:18)(cid:21)(cid:1) +<br/>% (cid:10)(cid:9)(cid:13)(cid:8)(cid:2)(cid:5)&(cid:11)(cid:5)(cid:19)(cid:4)(cid:6) ’((cid:6) +<br/>(cid:10)(cid:3)(cid:11)(cid:3)(cid:12)(cid:3)(cid:13)(cid:9) +<br/>(cid:1)(cid:2)(cid:3)(cid:4)(cid:5)(cid:3)(cid:4)(cid:6) (cid:7)(cid:3)(cid:8)(cid:9)(cid:6) +<br/>’(cid:16)(cid:4)(cid:1) (cid:26)(cid:20)(cid:8)(cid:12)(cid:6)(cid:8)(cid:12)(cid:1) 3(cid:8)(cid:27)(cid:4)(cid:1) (cid:19)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:14)(cid:1) (cid:24)(cid:16)(cid:4)(cid:1) (cid:25)(cid:6)(cid:20)(cid:11)(cid:24)(cid:1) (cid:6)(cid:17)(cid:8)-(cid:4)(cid:1) (cid:18)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:1) (cid:6)(cid:12)(cid:1) +<br/>(cid:17)(cid:6)(cid:18)(cid:18)(cid:5)(cid:4)E(cid:4)(cid:8)(cid:11)(cid:24)(cid:14)(cid:1)(cid:27)(cid:15)(cid:12)(cid:24)(cid:8)(cid:6)(cid:12)(cid:11)(cid:1)(cid:27)(cid:15)(cid:5)(cid:15)(cid:20)(cid:1)(cid:25)(cid:8)(cid:27)(cid:6)(cid:8)(cid:5)(cid:1)(cid:6)(cid:17)(cid:8)-(cid:4)(cid:20)(cid:30)(cid:1)(cid:15)(cid:25)(cid:1)(cid:8)(cid:1)(cid:5)(cid:8)(cid:20)-(cid:4)(cid:1)(cid:12))(cid:17)(cid:10)(cid:4)(cid:20)(cid:1)(cid:15)(cid:25)(cid:1) +<br/>(cid:26)(cid:20)(cid:8)(cid:12)(cid:6)(cid:8)(cid:12)(cid:1)(cid:11))(cid:10) (cid:4)(cid:27)(cid:24)(cid:11) (cid:10)(cid:4)(cid:24).(cid:4)(cid:4)(cid:12)(cid:1)/(cid:1)(cid:8)(cid:12)(cid:18)(cid:1)01(cid:1)(cid:30)(cid:4)(cid:8)(cid:20)(cid:11)(cid:1)(cid:15)(cid:5)(cid:18)(cid:21) +<br/>(cid:26)3(cid:19)(cid:23)(cid:1)(cid:6)(cid:11)(cid:1)(cid:8)(cid:1)(cid:5)(cid:8)(cid:20)-(cid:4)(cid:1)(cid:18)(cid:8)(cid:24)(cid:8)(cid:10)(cid:8)(cid:11)(cid:4)(cid:1)(cid:24)(cid:16)(cid:8)(cid:24)(cid:1)(cid:27)(cid:8)(cid:12)(cid:1)(cid:11))(((cid:15)(cid:20)(cid:24)(cid:1)(cid:11)(cid:24))(cid:18)(cid:6)(cid:4)(cid:11)(cid:1)(cid:15)(cid:25)(cid:1)(cid:24)(cid:16)(cid:4)(cid:1)(cid:8)-(cid:4)(cid:1) +<br/>(cid:27)(cid:5)(cid:8)(cid:11)(cid:11)(cid:6)(cid:25)(cid:6)(cid:27)(cid:8)(cid:24)(cid:6)(cid:15)(cid:12)(cid:1) (cid:11)(cid:30)(cid:11)(cid:24)(cid:4)(cid:17)(cid:11)(cid:21)(cid:1) (cid:26)(cid:24)(cid:1) (cid:27)(cid:15)(cid:12)(cid:24)(cid:8)(cid:6)(cid:12)(cid:11)(cid:1) (cid:15)(cid:29)(cid:4)(cid:20)(cid:1) *(cid:14)+,,(cid:1) (cid:27)(cid:15)(cid:5)(cid:15)(cid:20)(cid:1) (cid:6)(cid:17)(cid:8)-(cid:4)(cid:11)(cid:1) +<br/>(cid:13)(cid:15)(cid:1)(cid:20)(cid:4)(cid:11)(cid:24)(cid:20)(cid:6)(cid:27)(cid:24)(cid:6)(cid:15)(cid:12)(cid:11)(cid:1)(cid:15)(cid:12)(cid:1).(cid:4)(cid:8)(cid:20)(cid:1)4(cid:27)(cid:5)(cid:15)(cid:24)(cid:16)(cid:4)(cid:11)(cid:14)(cid:1)-(cid:5)(cid:8)(cid:11)(cid:11)(cid:4)(cid:11)(cid:14)(cid:1)(cid:4)(cid:24)(cid:27)(cid:21)5(cid:14)(cid:1) (cid:17)(cid:8)(cid:7)(cid:4)E)((cid:14)(cid:1)(cid:16)(cid:8)(cid:6)(cid:20)(cid:1) +<br/>(cid:11)(cid:24)(cid:30)(cid:5)(cid:4)(cid:14)(cid:1) (cid:25)(cid:8)(cid:27)(cid:6)(cid:8)(cid:5)(cid:1) (cid:16)(cid:8)(cid:6)(cid:20)(cid:1) .(cid:4)(cid:20)(cid:4)(cid:1) (cid:6)(cid:17)((cid:15)(cid:11)(cid:4)(cid:18)(cid:1) (cid:24)(cid:15)(cid:1) ((cid:8)(cid:20)(cid:24)(cid:6)(cid:27)(cid:6)((cid:8)(cid:12)(cid:24)(cid:11)(cid:21)(cid:1) D(cid:20)(cid:15))(cid:12)(cid:18)E(cid:24)(cid:20))(cid:24)(cid:16)(cid:1) +<br/>(cid:6)(cid:12)(cid:25)(cid:15)(cid:20)(cid:17)(cid:8)(cid:24)(cid:6)(cid:15)(cid:12)(cid:14)(cid:1)(cid:6)(cid:12)(cid:27)(cid:5))(cid:18)(cid:6)(cid:12)-(cid:1)(cid:26)(cid:19)(cid:14)(cid:1)(cid:8)-(cid:4)(cid:14)(cid:1)(cid:7)(cid:6)(cid:12)(cid:18)(cid:1)(cid:15)(cid:25) ((cid:15)(cid:11)(cid:4)(cid:1)(cid:15)(cid:20)(cid:1)(cid:4)6((cid:20)(cid:4)(cid:11)(cid:11)(cid:6)(cid:15)(cid:12)(cid:1)(cid:8)(cid:12)(cid:18)(cid:1) +<br/>(cid:6)(cid:25)(cid:1) (cid:24)(cid:16)(cid:4)(cid:1) (cid:11))(cid:10) (cid:4)(cid:27)(cid:24)(cid:1) (cid:16)(cid:8)(cid:11)(cid:1) -(cid:5)(cid:8)(cid:11)(cid:11)(cid:4)(cid:11)(cid:1) (cid:6)(cid:11)(cid:1) ((cid:20)(cid:15)(cid:29)(cid:6)(cid:18)(cid:4)(cid:18)(cid:21)(cid:1) #6((cid:4)(cid:20)(cid:6)(cid:17)(cid:4)(cid:12)(cid:24)(cid:8)(cid:5)(cid:1) (cid:11))(cid:10) (cid:4)(cid:27)(cid:24)(cid:11)(cid:1) +<br/>.(cid:4)(cid:20)(cid:4)(cid:1)((cid:16)(cid:15)(cid:24)(cid:15)-(cid:20)(cid:8)((cid:16)(cid:4)(cid:18)(cid:1).(cid:6)(cid:24)(cid:16)(cid:1)(cid:8)(cid:1)(cid:25)(cid:6)(cid:12)(cid:4)E(cid:20)(cid:4)(cid:11)(cid:15)(cid:5))(cid:24)(cid:6)(cid:15)(cid:12)(cid:1)(cid:27)(cid:15)(cid:5)(cid:15)(cid:20)(cid:1)(cid:18)(cid:6)-(cid:6)(cid:24)(cid:8)(cid:5)(cid:1)(cid:27)(cid:8)(cid:17)(cid:4)(cid:20)(cid:8)(cid:1) +<br/>(cid:6)(cid:12)(cid:1)(cid:18)(cid:8)(cid:30)(cid:5)(cid:6)-(cid:16)(cid:24)(cid:21)(cid:1)’(cid:16)(cid:4)(cid:1)(cid:11))(cid:10) (cid:4)(cid:27)(cid:24)(cid:11)(cid:1).(cid:4)(cid:20)(cid:4)(cid:1)(cid:11)(cid:4)(cid:8)(cid:24)(cid:4)(cid:18)(cid:1)(cid:15)(cid:12)(cid:1)(cid:8)(cid:1)(cid:11)(cid:24)(cid:15)(cid:15)(cid:5)(cid:1)(cid:8)(cid:12)(cid:18)(cid:1)(cid:6)(cid:12)(cid:11)(cid:24)(cid:20))(cid:27)(cid:24)(cid:4)(cid:18)(cid:1) +<br/>(cid:24)(cid:15)(cid:1) (cid:17)(cid:8)(cid:6)(cid:12)(cid:24)(cid:8)(cid:6)(cid:12)(cid:1) (cid:8)(cid:1) (cid:27)(cid:15)(cid:12)(cid:11)(cid:24)(cid:8)(cid:12)(cid:24)(cid:1) (cid:16)(cid:4)(cid:8)(cid:18)(cid:1) ((cid:15)(cid:11)(cid:6)(cid:24)(cid:6)(cid:15)(cid:12)(cid:1) 4(cid:8)(cid:5)(cid:24)(cid:16)(cid:15))-(cid:16)(cid:1) (cid:11)(cid:5)(cid:6)-(cid:16)(cid:24)(cid:1) +<br/>(cid:17)(cid:15)(cid:29)(cid:4)(cid:17)(cid:4)(cid:12)(cid:24)(cid:11)(cid:1).(cid:4)(cid:20)(cid:4)(cid:1))(cid:12)(cid:8)(cid:29)(cid:15)(cid:6)(cid:18)(cid:8)(cid:10)(cid:5)(cid:4)5(cid:21) +<br/>’(cid:16)(cid:4)(cid:1)(cid:6)(cid:17)(cid:8)-(cid:4)(cid:11)(cid:1)(cid:8)(cid:20)(cid:4)(cid:1)(cid:6)(cid:12)(cid:1)>0,G+>,(cid:1)((cid:6)6(cid:4)(cid:5)(cid:11)(cid:1)(cid:20)(cid:4)(cid:11)(cid:15)(cid:5))(cid:24)(cid:6)(cid:15)(cid:12)(cid:14)(cid:1)/>(cid:1)(cid:10)(cid:6)(cid:24)(cid:1)(cid:18)(cid:4)((cid:24)(cid:16) (cid:14)(cid:1) +<br/>(cid:8)(cid:10)(cid:15))(cid:24)(cid:1)>,(cid:1)(cid:31)(cid:10)(cid:30)(cid:24)(cid:4)(cid:11)(cid:1)(cid:11)(cid:6)(cid:22)(cid:4)(cid:1)(cid:8)(cid:12)(cid:18)(cid:1)CHD(cid:1)(cid:25)(cid:15)(cid:20)(cid:17)(cid:8)(cid:24) (cid:21)(cid:1) +<br/>#(cid:12)(cid:15))-(cid:16)(cid:1) (cid:5))(cid:17)(cid:6)(cid:12)(cid:15)(cid:11)(cid:6)(cid:24)(cid:30)(cid:1) (cid:25)(cid:15)(cid:20)(cid:1) .(cid:20)(cid:6)(cid:12)(cid:7)(cid:5)(cid:4)(cid:1) ((cid:20)(cid:15)(cid:27)(cid:4)(cid:11)(cid:11)(cid:6)(cid:12)-(cid:1) (cid:8)(cid:12)(cid:18)(cid:1) (cid:25)(cid:8)(cid:27)(cid:6)(cid:8)(cid:5)(cid:1) +<br/>(cid:25)(cid:4)(cid:8)(cid:24))(cid:20)(cid:4)(cid:11)(cid:1) .(cid:6)(cid:24)(cid:16)(cid:15))(cid:24)(cid:1) (cid:11)(cid:16)(cid:8)(cid:18)(cid:15).(cid:11)(cid:1) (cid:6)(cid:11)(cid:1) (cid:12)(cid:4)(cid:4)(cid:18)(cid:4)(cid:18)(cid:1) 4(cid:6)(cid:12)(cid:1) (cid:8)-(cid:4)(cid:1) (cid:27)(cid:5)(cid:8)(cid:11)(cid:11)(cid:6)(cid:25)(cid:6)(cid:27)(cid:8)(cid:24)(cid:6)(cid:15)(cid:12)(cid:1) +<br/>.(cid:20)(cid:6)(cid:12)(cid:7)(cid:5)(cid:4)(cid:1) (cid:18)(cid:4)(cid:24)(cid:4)(cid:27)(cid:24)(cid:6)(cid:15)(cid:12)(cid:1) (cid:8)(cid:12)(cid:18)(cid:1) (cid:8)(cid:12)(cid:8)(cid:5)(cid:30)(cid:11)(cid:6)(cid:11)(cid:1) +<br/>(cid:24)(cid:16)(cid:4)(cid:1) +<br/>(cid:18)(cid:6)(cid:11)(cid:24)(cid:6)(cid:12)-)(cid:6)(cid:11)(cid:16)(cid:6)(cid:12)-(cid:1)(cid:15)(cid:25)(cid:1)(cid:11)(cid:4)(cid:12)(cid:6)(cid:15)(cid:20)(cid:11)(cid:1)(cid:25)(cid:20)(cid:15)(cid:17)(cid:1)(cid:24)(cid:16)(cid:15)(cid:11)(cid:4)(cid:1)(cid:6)(cid:12)(cid:1)(cid:24)(cid:16)(cid:4)(cid:1)(cid:30)(cid:15))(cid:12)-(cid:4)(cid:20)(cid:1)(cid:27)(cid:8)(cid:24)(cid:4)-(cid:15)(cid:20)(cid:6)(cid:4)(cid:11)(cid:1) +<br/><(cid:2)=5(cid:21)(cid:1) ")(cid:10) (cid:4)(cid:27)(cid:24)(cid:11)(cid:1) .(cid:4)(cid:20)(cid:4)(cid:1) ((cid:16)(cid:15)(cid:24)(cid:15)-(cid:20)(cid:8)((cid:16)(cid:4)(cid:18)(cid:1) .(cid:6)(cid:24)(cid:16)(cid:15))(cid:24)(cid:1) (cid:8)(cid:12)(cid:30)(cid:1) ((cid:20)(cid:15) (cid:4)(cid:27)(cid:24)(cid:15)(cid:20)(cid:11)(cid:1) (cid:15)(cid:20)(cid:1) +<br/>(cid:6)(cid:17)((cid:15)(cid:20)(cid:24)(cid:8)(cid:12)(cid:24)(cid:1) +<br/>(cid:25)(cid:15)(cid:20)(cid:1) +<br/>(cid:6)(cid:11)(cid:1) +</td></tr><tr><td>06526c52a999fdb0a9fd76e84f9795a69480cecf</td><td></td></tr><tr><td>06fe63b34fcc8ff68b72b5835c4245d3f9b8a016</td><td>Mach Learn +<br/>DOI 10.1007/s10994-013-5336-9 +<br/>Learning semantic representations of objects +<br/>and their parts +<br/>Received: 24 May 2012 / Accepted: 26 February 2013 +<br/>© The Author(s) 2013 +</td></tr><tr><td>06aab105d55c88bd2baa058dc51fa54580746424</td><td>Image Set based Collaborative Representation for +<br/>Face Recognition +</td></tr><tr><td>06262d14323f9e499b7c6e2a3dec76ad9877ba04</td><td>Real-Time Pose Estimation Piggybacked on Object Detection +<br/>Brno, Czech Republic +</td></tr><tr><td>062c41dad67bb68fefd9ff0c5c4d296e796004dc</td><td>Temporal Generative Adversarial Nets with Singular Value Clipping +<br/>Preferred Networks inc., Japan +</td></tr><tr><td>06400a24526dd9d131dfc1459fce5e5189b7baec</td><td>Event Recognition in Photo Collections with a Stopwatch HMM +<br/>1Computer Vision Lab +<br/>ETH Z¨urich, Switzerland +<br/>2ESAT, PSI-VISICS +<br/>K.U. Leuven, Belgium +</td></tr><tr><td>0653dcdff992ad980cd5ea5bc557efb6e2a53ba1</td><td></td></tr><tr><td>063a3be18cc27ba825bdfb821772f9f59038c207</td><td>This is a repository copy of The development of spontaneous facial responses to others’ +<br/>emotions in infancy. An EMG study. +<br/>White Rose Research Online URL for this paper: +<br/>http://eprints.whiterose.ac.uk/125231/ +<br/>Version: Published Version +<br/>Article: +<br/>Kaiser, Jakob, Crespo-Llado, Maria Magdalena, Turati, Chiara et al. (1 more author) +<br/>(2017) The development of spontaneous facial responses to others’ emotions in infancy. +<br/>An EMG study. Scientific Reports. ISSN 2045-2322 +<br/>https://doi.org/10.1038/s41598-017-17556-y +<br/>Reuse +<br/>This article is distributed under the terms of the Creative Commons Attribution (CC BY) licence. This licence +<br/>allows you to distribute, remix, tweak, and build upon the work, even commercially, as long as you credit the +<br/>authors for the original work. More information and the full terms of the licence here: +<br/>https://creativecommons.org/licenses/ +<br/>Takedown +<br/>If you consider content in White Rose Research Online to be in breach of UK law, please notify us by +<br/>https://eprints.whiterose.ac.uk/ +</td></tr><tr><td>06a9ed612c8da85cb0ebb17fbe87f5a137541603</td><td>Deep Learning of Player Trajectory Representations for Team +<br/>Activity Analysis +</td></tr><tr><td>06ad99f19cf9cb4a40741a789e4acbf4433c19ae</td><td>SenTion: A framework for Sensing Facial +<br/>Expressions +</td></tr><tr><td>6c304f3b9c3a711a0cca5c62ce221fb098dccff0</td><td>Attentive Semantic Video Generation using Captions +<br/>IIT Hyderabad +<br/>IIT Hyderabad +</td></tr><tr><td>6c2b392b32b2fd0fe364b20c496fcf869eac0a98</td><td>DOI 10.1007/s00138-012-0423-7 +<br/>ORIGINAL PAPER +<br/>Fully automatic face recognition framework based +<br/>on local and global features +<br/>Received: 30 May 2011 / Revised: 21 February 2012 / Accepted: 29 February 2012 / Published online: 22 March 2012 +<br/>© Springer-Verlag 2012 +</td></tr><tr><td>6cddc7e24c0581c50adef92d01bb3c73d8b80b41</td><td>Face Verification Using the LARK +<br/>Representation +</td></tr><tr><td>6c8c7065d1041146a3604cbe15c6207f486021ba</td><td>Attention Modeling for Face Recognition via Deep Learning +<br/>Department of Computing, Hung Hom, Kowloon +<br/>Hong Kong, 999077 CHINA +<br/>Department of Computing, Hung Hom, Kowloon +<br/>Hong Kong, 99907 CHINA +<br/>Department of Computing, Hung Hom, Kowloon +<br/>Hong Kong, 99907 CHINA +<br/>Department of Computing, Hung Hom, Kowloon +<br/>Hong Kong, 99907 CHINA +</td></tr><tr><td>390f3d7cdf1ce127ecca65afa2e24c563e9db93b</td><td>Learning Deep Representation for Face +<br/>Alignment with Auxiliary Attributes +</td></tr><tr><td>3918b425bb9259ddff9eca33e5d47bde46bd40aa</td><td>Copyright +<br/>by +<br/>David Lieh-Chiang Chen +<br/>2012 +</td></tr><tr><td>39ce143238ea1066edf0389d284208431b53b802</td><td></td></tr><tr><td>39ce2232452c0cd459e32a19c1abe2a2648d0c3f</td><td></td></tr><tr><td>3998c5aa6be58cce8cb65a64cb168864093a9a3e</td><td></td></tr><tr><td>397aeaea61ecdaa005b09198942381a7a11cd129</td><td></td></tr><tr><td>39b22bcbd452d5fea02a9ee63a56c16400af2b83</td><td></td></tr><tr><td>399a2c23bd2592ebe20aa35a8ea37d07c14199da</td><td></td></tr><tr><td>39c8b34c1b678235b60b648d0b11d241a34c8e32</td><td>Learning to Deblur Images with Exemplars +</td></tr><tr><td>3986161c20c08fb4b9b791b57198b012519ea58b</td><td>International Journal of Soft Computing and Engineering (IJSCE) +<br/>ISSN: 2231-2307, Volume-4 Issue-4, September 2014 +<br/>An Efficient Method for Face Recognition based on +<br/>Fusion of Global and Local Feature Extraction +</td></tr><tr><td>392425be1c9d9c2ee6da45de9df7bef0d278e85f</td><td></td></tr><tr><td>392c3cabe516c0108b478152902a9eee94f4c81e</td><td>Computer Science and Artificial Intelligence Laboratory +<br/>Technical Report +<br/>MIT-CSAIL-TR-2007-024 +<br/>April 23, 2007 +<br/>Tiny images +<br/>m a s s a c h u s e t t s i n s t i t u t e o f t e c h n o l o g y, c a m b r i d g e , m a 0 213 9 u s a — w w w. c s a i l . m i t . e d u +</td></tr><tr><td>3947b64dcac5bcc1d3c8e9dcb50558efbb8770f1</td><td></td></tr><tr><td>3965d61c4f3b72044f43609c808f8760af8781a2</td><td></td></tr><tr><td>395bf182983e0917f33b9701e385290b64e22f9a</td><td></td></tr><tr><td>3933e323653ff27e68c3458d245b47e3e37f52fd</td><td>Evaluation of a 3D-aided Pose Invariant 2D Face Recognition System +<br/>Computational Biomedicine Lab +<br/>4800 Calhoun Rd. Houston, TX, USA +</td></tr><tr><td>39b452453bea9ce398613d8dd627984fd3a0d53c</td><td></td></tr><tr><td>3958db5769c927cfc2a9e4d1ee33ecfba86fe054</td><td>Describable Visual Attributes for +<br/>Face Verification and Image Search +</td></tr><tr><td>39b5f6d6f8d8127b2b97ea1a4987732c0db6f9df</td><td></td></tr><tr><td>994f7c469219ccce59c89badf93c0661aae34264</td><td>1 +<br/>Model Based Face Recognition Across Facial +<br/>Expressions +<br/> +<br/>screens, embedded into mobiles and installed into everyday +<br/>living and working environments they become valuable tools +<br/>for human system interaction. A particular important aspect of +<br/>this interaction is detection and recognition of faces and +<br/>interpretation of facial expressions. These capabilities are +<br/>deeply rooted in the human visual system and a crucial +<br/>building block for social interaction. Consequently, these +<br/>capabilities are an important step towards the acceptance of +<br/>many technical systems. +<br/>trees as a classifier +<br/>lies not only +</td></tr><tr><td>9949ac42f39aeb7534b3478a21a31bc37fe2ffe3</td><td>Parametric Stereo for Multi-Pose Face Recognition and +<br/>3D-Face Modeling +<br/>PSI ESAT-KUL +<br/>Leuven, Belgium +</td></tr><tr><td>9958942a0b7832e0774708a832d8b7d1a5d287ae</td><td>The Sparse Matrix Transform for Covariance +<br/>Estimation and Analysis of High Dimensional +<br/>Signals +</td></tr><tr><td>9931c6b050e723f5b2a189dd38c81322ac0511de</td><td></td></tr><tr><td>9993f1a7cfb5b0078f339b9a6bfa341da76a3168</td><td>JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 +<br/>A Simple, Fast and Highly-Accurate Algorithm to +<br/>Recover 3D Shape from 2D Landmarks on a Single +<br/>Image +</td></tr><tr><td>99c20eb5433ed27e70881d026d1dbe378a12b342</td><td>ISCA Archive +<br/>http://www.isca-speech.org/archive +<br/>First Workshop on Speech, Language +<br/>and Audio in Multimedia +<br/>Marseille, France +<br/>August 22-23, 2013 +<br/>Proceedings of the First Workshop on Speech, Language and Audio in Multimedia (SLAM), Marseille, France, August 22-23, 2013. +<br/>78 +</td></tr><tr><td>9990e0b05f34b586ffccdc89de2f8b0e5d427067</td><td>International Journal of Modeling and Optimization, Vol. 3, No. 2, April 2013 +<br/>Auto-Optimized Multimodal Expression Recognition +<br/>Framework Using 3D Kinect Data for ASD Therapeutic +<br/>Aid +<br/> +<br/>regarding +<br/>emotion +<br/>and +<br/>to +<br/>recognize +</td></tr><tr><td>99d7678039ad96ee29ab520ff114bb8021222a91</td><td>Political image analysis with deep neural +<br/>networks +<br/>November 28, 2017 +</td></tr><tr><td>529e2ce6fb362bfce02d6d9a9e5de635bde81191</td><td>This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. +<br/>> TIP-05732-2009< +<br/>1 +<br/>Normalization of Face Illumination Based +<br/>on Large- and Small- Scale Features +</td></tr><tr><td>52887969107956d59e1218abb84a1f834a314578</td><td>1283 +<br/>Travel Recommendation by Mining People +<br/>Attributes and Travel Group Types From +<br/>Community-Contributed Photos +</td></tr><tr><td>521482c2089c62a59996425603d8264832998403</td><td></td></tr><tr><td>521b625eebea73b5deb171a350e3709a4910eebf</td><td></td></tr><tr><td>527dda77a3864d88b35e017d542cb612f275a4ec</td><td></td></tr><tr><td>52f23e1a386c87b0dab8bfdf9694c781cd0a3984</td><td></td></tr><tr><td>529baf1a79cca813f8c9966ceaa9b3e42748c058</td><td>Triangle Wise Mapping Technique to Transform one Face Image into Another Face Image +<br/> +<br/>{tag} {/tag} +<br/> +<br/> International Journal of Computer Applications +<br/> +<br/> © 2014 by IJCA Journal +<br/> Volume 87 - Number 6 +<br/> +<br/> Year of Publication: 2014 +<br/> +<br/> +<br/> +<br/> Authors: +<br/> +<br/>Bhogeswar Borah +<br/> +<br/> +<br/> +<br/> +<br/> +<br/> +<br/> +<br/> +<br/> +<br/> +<br/> 10.5120/15209-3714 +<br/> {bibtex}pxc3893714.bib{/bibtex} +</td></tr><tr><td>5239001571bc64de3e61be0be8985860f08d7e7e</td><td>SUBMITTED TO IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, JUNE 2016 +<br/>Deep Appearance Models: A Deep Boltzmann +<br/>Machine Approach for Face Modeling +</td></tr><tr><td>550858b7f5efaca2ebed8f3969cb89017bdb739f</td><td></td></tr><tr><td>554b9478fd285f2317214396e0ccd81309963efd</td><td>Spatio-Temporal Action Localization For Human Action +<br/>Recognition in Large Dataset +<br/>1L2TI, Institut Galil´ee, Universit´e Paris 13, France; +<br/>2SERCOM, Ecole Polytechnique de Tunisie +</td></tr><tr><td>55c68c1237166679d2cb65f266f496d1ecd4bec6</td><td>Learning to Score Figure Skating Sport Videos +</td></tr><tr><td>5502dfe47ac26e60e0fb25fc0f810cae6f5173c0</td><td>Affordance Prediction via Learned Object Attributes +</td></tr><tr><td>55a158f4e7c38fe281d06ae45eb456e05516af50</td><td>The 22nd International Conference on Computer Graphics and Vision +<br/>108 +<br/>GraphiCon’2012 +</td></tr><tr><td>5506a1a1e1255353fde05d9188cb2adc20553af5</td><td></td></tr><tr><td>55c81f15c89dc8f6eedab124ba4ccab18cf38327</td><td></td></tr><tr><td>551fa37e8d6d03b89d195a5c00c74cc52ff1c67a</td><td>GeThR-Net: A Generalized Temporally Hybrid +<br/>Recurrent Neural Network for Multimodal +<br/>Information Fusion +<br/>1 Xerox Research Centre India; 2 Amazon Development Center India +</td></tr><tr><td>55c40cbcf49a0225e72d911d762c27bb1c2d14aa</td><td>Indian Face Age Database: A Database for Face Recognition with Age Variation +<br/>{tag} {/tag} +<br/> International Journal of Computer Applications +<br/> +<br/> Foundation of Computer Science (FCS), NY, USA +<br/> +<br/> +<br/>Volume 126 +<br/>- +<br/>Number 5 +<br/> +<br/> +<br/> Year of Publication: 2015 +<br/> +<br/> +<br/> +<br/> +<br/> Authors: +<br/> +<br/> +<br/> +<br/> +<br/> +<br/> +<br/> +<br/> +<br/> +<br/> +<br/> +<br/> 10.5120/ijca2015906055 +<br/> {bibtex}2015906055.bib{/bibtex} +</td></tr><tr><td>973e3d9bc0879210c9fad145a902afca07370b86</td><td>(IJACSA) International Journal of Advanced Computer Science and Applications, +<br/>Vol. 7, No. 7, 2016 +<br/>From Emotion Recognition to Website +<br/>Customizations +<br/>O.B. Efremides +<br/>School of Web Media +<br/>Bahrain Polytechnic +<br/>Isa Town, Kingdom of Bahrain +</td></tr><tr><td>97b8249914e6b4f8757d22da51e8347995a40637</td><td>28 +<br/>Large-Scale Vehicle Detection, Indexing, +<br/>and Search in Urban Surveillance Videos +</td></tr><tr><td>97032b13f1371c8a813802ade7558e816d25c73f</td><td>Total Recall Final Report +<br/>Supervisor: Professor Duncan Gillies +<br/>January 11, 2006 +</td></tr><tr><td>97cf04eaf1fc0ac4de0f5ad4a510d57ce12544f5</td><td>manuscript No. +<br/>(will be inserted by the editor) +<br/>Deep Affect Prediction in-the-wild: Aff-Wild Database and Challenge, +<br/>Deep Architectures, and Beyond +<br/>Zafeiriou4 +</td></tr><tr><td>97d1d561362a8b6beb0fdbee28f3862fb48f1380</td><td>1955 +<br/>Age Synthesis and Estimation via Faces: +<br/>A Survey +</td></tr><tr><td>97540905e4a9fdf425989a794f024776f28a3fa9</td><td></td></tr><tr><td>9755554b13103df634f9b1ef50a147dd02eab02f</td><td>How Transferable are CNN-based Features for +<br/>Age and Gender Classification? +<br/> 1 +</td></tr><tr><td>635158d2da146e9de559d2742a2fa234e06b52db</td><td></td></tr><tr><td>63cf5fc2ee05eb9c6613043f585dba48c5561192</td><td>Prototype Selection for +<br/>Classification in Standard +<br/>and Generalized +<br/>Dissimilarity Spaces +</td></tr><tr><td>63d8d69e90e79806a062cb8654ad78327c8957bb</td><td></td></tr><tr><td>631483c15641c3652377f66c8380ff684f3e365c</td><td>Sync-DRAW: Automatic Video Generation using Deep Recurrent +<br/>A(cid:130)entive Architectures +<br/>Gaurav Mi(cid:138)al∗ +<br/>IIT Hyderabad +<br/>Vineeth N Balasubramanian +<br/>IIT Hyderabad +</td></tr><tr><td>63eefc775bcd8ccad343433fc7a1dd8e1e5ee796</td><td></td></tr><tr><td>632fa986bed53862d83918c2b71ab953fd70d6cc</td><td>GÜNEL ET AL.: WHAT FACE AND BODY SHAPES CAN TELL ABOUT HEIGHT +<br/>What Face and Body Shapes Can Tell +<br/>About Height +<br/>CVLab +<br/>EPFL, +<br/>Lausanne, Switzerland +</td></tr><tr><td>63340c00896d76f4b728dbef85674d7ea8d5ab26</td><td>1732 +<br/>Discriminant Subspace Analysis: +<br/>A Fukunaga-Koontz Approach +</td></tr><tr><td>63d865c66faaba68018defee0daf201db8ca79ed</td><td>Deep Regression for Face Alignment +<br/>1Dept. of Electronics and Information Engineering, Huazhong Univ. of Science and Technology, China +<br/>2Microsoft Research, Beijing, China +</td></tr><tr><td>634541661d976c4b82d590ef6d1f3457d2857b19</td><td>AAllmmaa MMaatteerr SSttuuddiioorruumm –– UUnniivveerrssiittàà ddii BBoollooggnnaa +<br/>in cotutela con Università di Sassari +<br/>DOTTORATO DI RICERCA IN +<br/>INGEGNERIA ELETTRONICA, INFORMATICA E DELLE +<br/>TELECOMUNICAZIONI +<br/>Ciclo XXVI +<br/>Settore Concorsuale di afferenza: 09/H1 +<br/>Settore Scientifico disciplinare: ING-INF/05 +<br/>ADVANCED TECHNIQUES FOR FACE RECOGNITION +<br/>UNDER CHALLENGING ENVIRONMENTS +<br/>TITOLO TESI +<br/>Presentata da: +<br/>Coordinatore Dottorato +<br/>ALESSANDRO VANELLI-CORALLI +<br/> +<br/>Relatore +<br/> DAVIDE MALTONI +<br/>Relatore +<br/> MASSIMO TISTARELLI +<br/>Esame finale anno 2014 +</td></tr><tr><td>6332a99e1680db72ae1145d65fa0cccb37256828</td><td>MASTER IN COMPUTER VISION AND ARTIFICIAL INTELLIGENCE +<br/>REPORT OF THE RESEARCH PROJECT +<br/>OPTION: COMPUTER VISION +<br/>Pose and Face Recovery via +<br/>Spatio-temporal GrabCut Human +<br/>Segmentation +<br/>Date: 13/07/2010 +</td></tr><tr><td>63c022198cf9f084fe4a94aa6b240687f21d8b41</td><td>425 +</td></tr><tr><td>0f65c91d0ed218eaa7137a0f6ad2f2d731cf8dab</td><td>Multi-Directional Multi-Level Dual-Cross +<br/>Patterns for Robust Face Recognition +</td></tr><tr><td>0f112e49240f67a2bd5aaf46f74a924129f03912</td><td>947 +<br/>Age-Invariant Face Recognition +</td></tr><tr><td>0f4cfcaca8d61b1f895aa8c508d34ad89456948e</td><td>LOCAL APPEARANCE BASED FACE RECOGNITION USING +<br/>DISCRETE COSINE TRANSFORM (WedPmPO4) +<br/>Author(s) : +</td></tr><tr><td>0fad544edfc2cd2a127436a2126bab7ad31ec333</td><td>Decorrelating Semantic Visual Attributes by Resisting the Urge to Share +<br/>UT Austin +<br/>USC +<br/>UT Austin +</td></tr><tr><td>0f32df6ae76402b98b0823339bd115d33d3ec0a0</td><td>Emotion recognition from embedded bodily +<br/>expressions and speech during dyadic interactions +</td></tr><tr><td>0fd1715da386d454b3d6571cf6d06477479f54fc</td><td>J Intell Robot Syst (2016) 82:101–133 +<br/>DOI 10.1007/s10846-015-0259-2 +<br/>A Survey of Autonomous Human Affect Detection Methods +<br/>for Social Robots Engaged in Natural HRI +<br/>Received: 10 December 2014 / Accepted: 11 August 2015 / Published online: 23 August 2015 +<br/>© Springer Science+Business Media Dordrecht 2015 +</td></tr><tr><td>0f9bf5d8f9087fcba419379600b86ae9e9940013</td><td></td></tr><tr><td>0f92e9121e9c0addc35eedbbd25d0a1faf3ab529</td><td>MORPH-II: A Proposed Subsetting Scheme +<br/>NSF-REU Site at UNC Wilmington, Summer 2017 +</td></tr><tr><td>0fd1bffb171699a968c700f206665b2f8837d953</td><td>Weakly Supervised Object Localization with +<br/>Multi-fold Multiple Instance Learning +</td></tr><tr><td>0a511058edae582e8327e8b9d469588c25152dc6</td><td></td></tr><tr><td>0a4f3a423a37588fde9a2db71f114b293fc09c50</td><td></td></tr><tr><td>0a3863a0915256082aee613ba6dab6ede962cdcd</td><td>Early and Reliable Event Detection Using Proximity Space Representation +<br/>LTCI, CNRS, T´el´ecom ParisTech, Universit´e Paris-Saclay, 75013, Paris, France +<br/>J´erˆome Gauthier +<br/>LADIS, CEA, LIST, 91191, Gif-sur-Yvette, France +<br/>Normandie Universit´e, UR, LITIS EA 4108, Avenue de l’universit´e, 76801, Saint-Etienne-du-Rouvray, France +</td></tr><tr><td>0ad90118b4c91637ee165f53d557da7141c3fde0</td><td></td></tr><tr><td>0af48a45e723f99b712a8ce97d7826002fe4d5a5</td><td>2982 +<br/>Toward Wide-Angle Microvision Sensors +<br/>Todd Zickler, Member, IEEE +</td></tr><tr><td>0aa8a0203e5f406feb1815f9b3dd49907f5fd05b</td><td>Mixture subclass discriminant analysis +</td></tr><tr><td>0a1138276c52c734b67b30de0bf3f76b0351f097</td><td>This is the author's version of an article that has been published in this journal. Changes were made to this version by the publisher prior to publication. +<br/>The final version of record is available at +<br/> http://dx.doi.org/10.1109/TIP.2016.2539502 +<br/>Discriminant Incoherent Component Analysis +</td></tr><tr><td>0a6a25ee84fc0bf7284f41eaa6fefaa58b5b329a</td><td></td></tr><tr><td>0ae9cc6a06cfd03d95eee4eca9ed77b818b59cb7</td><td>Noname manuscript No. +<br/>(will be inserted by the editor) +<br/>Multi-task, multi-label and multi-domain learning with +<br/>residual convolutional networks for emotion recognition +<br/>Received: date / Accepted: date +</td></tr><tr><td>0acf23485ded5cb9cd249d1e4972119239227ddb</td><td>Dual coordinate solvers for large-scale structural SVMs +<br/>UC Irvine +<br/>This manuscript describes a method for training linear SVMs (including binary SVMs, SVM regression, +<br/>and structural SVMs) from large, out-of-core training datasets. Current strategies for large-scale learning fall +<br/>into one of two camps; batch algorithms which solve the learning problem given a finite datasets, and online +<br/>algorithms which can process out-of-core datasets. The former typically requires datasets small enough to fit +<br/>in memory. The latter is often phrased as a stochastic optimization problem [4, 15]; such algorithms enjoy +<br/>strong theoretical properties but often require manual tuned annealing schedules, and may converge slowly +<br/>for problems with large output spaces (e.g., structural SVMs). We discuss an algorithm for an “intermediate” +<br/>regime in which the data is too large to fit in memory, but the active constraints (support vectors) are small +<br/>enough to remain in memory. +<br/>In this case, one can design rather efficient learning algorithms that are +<br/>as stable as batch algorithms, but capable of processing out-of-core datasets. We have developed such a +<br/>MATLAB-based solver and used it to train a series of recognition systems [19, 7, 21, 12] for articulated pose +<br/>estimation, facial analysis, 3D object recognition, and action classification, all with publicly-available code. +<br/>This writeup describes the solver in detail. +<br/>Approach: Our approach is closely based on data-subsampling algorithms for collecting hard exam- +<br/>ples [9, 10, 6], combined with the dual coordinate quadratic programming (QP) solver described in liblinear +<br/>[8]. The latter appears to be current fastest method for learning linear SVMs. We make two extensions (1) +<br/>We show how to generalize the solver to other types of SVM problems such as (latent) structural SVMs (2) +<br/>We show how to modify it to behave as a partially-online algorithm, which only requires access to small +<br/>amounts of data at a time. +<br/>Overview: Sec. 1 describes a general formulation of an SVM problem that encompasses many standard +<br/>tasks such as multi-class classification and (latent) structural prediction. Sec. 2 derives its dual QP, and Sec. 3 +<br/>describes a dual coordinate descent optimization algorithm. Sec. 4 describes modifications for optimizing +<br/>in an online fashion, allowing one to learn near-optimal models with a single pass over large, out-of-core +<br/>datasets. Sec. 5 briefly touches on some theoretical issues that are necessary to ensure convergence. Finally, +<br/>Sec. 6 and Sec. 7 describe modifications to our basic formulation to accommodate non-negativity constraints +<br/>and flexible regularization schemes during learning. +<br/>1 Generalized SVMs +<br/>We first describe a general formulation of a SVM which encompasses various common problems such as +<br/>binary classification, regression, and structured prediction. Assume we are given training data where the ith +<br/>example is described by a set of Ni vectors {xij} and a set of Ni scalars {lij}, where j varies from 1 to Ni. +<br/>We wish to solve the following optimization problem: +<br/>(0, lij − wT xij) +<br/>max +<br/>j∈Ni +<br/>(1) +<br/>(cid:88) +<br/>argmin +<br/>L(w) = +<br/>||w||2 + +</td></tr><tr><td>0ad4a814b30e096ad0e027e458981f812c835aa0</td><td></td></tr><tr><td>6448d23f317babb8d5a327f92e199aaa45f0efdc</td><td></td></tr><tr><td>6412d8bbcc01f595a2982d6141e4b93e7e982d0f</td><td>Deep Convolutional Neural Network using Triplets of Faces, Deep Ensemble, and +<br/>Score-level Fusion for Face Recognition +<br/>1Department of Creative IT Engineering, POSTECH, Korea +<br/>2Department of Computer Science and Engineering, POSTECH, Korea +</td></tr><tr><td>649eb674fc963ce25e4e8ce53ac7ee20500fb0e3</td><td></td></tr><tr><td>642c66df8d0085d97dc5179f735eed82abf110d0</td><td></td></tr><tr><td>641f34deb3bdd123c6b6e7b917519c3e56010cb7</td><td></td></tr><tr><td>645de797f936cb19c1b8dba3b862543645510544</td><td>Deep Temporal Linear Encoding Networks +<br/>1ESAT-PSI, KU Leuven, 2CVL, ETH Z¨urich +</td></tr><tr><td>6462ef39ca88f538405616239471a8ea17d76259</td><td></td></tr><tr><td>90ac0f32c0c29aa4545ed3d5070af17f195d015f</td><td></td></tr><tr><td>90cb074a19c5e7d92a1c0d328a1ade1295f4f311</td><td>MIT. Media Laboratory Affective Computing Technical Report #571 +<br/>Appears in IEEE International Workshop on Analysis and Modeling of Faces and Gestures , Oct 2003 +<br/>Fully Automatic Upper Facial Action Recognition +<br/>MIT Media Laboratory +<br/>Cambridge, MA 02139 +</td></tr><tr><td>90b11e095c807a23f517d94523a4da6ae6b12c76</td><td></td></tr><tr><td>9028fbbd1727215010a5e09bc5758492211dec19</td><td>Solving the Uncalibrated Photometric Stereo +<br/>Problem using Total Variation +<br/>1 IRIT, UMR CNRS 5505, Toulouse, France +<br/>2 Dept. of Computer Science, Univ. of Copenhagen, Denmark +</td></tr><tr><td>bf1e0279a13903e1d43f8562aaf41444afca4fdc</td><td> International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 +<br/> Volume: 04 Issue: 10 | Oct -2017 www.irjet.net p-ISSN: 2395-0072 +<br/>Different Viewpoints of Recognizing Fleeting Facial Expressions with +<br/>DWT +<br/>information +<br/>to get desired +<br/>information +<br/>Introduction +<br/>---------------------------------------------------------------------***--------------------------------------------------------------------- +</td></tr><tr><td>bf5940d57f97ed20c50278a81e901ae4656f0f2c</td><td>Query-free Clothing Retrieval via Implicit +<br/>Relevance Feedback +</td></tr><tr><td>bfb98423941e51e3cd067cb085ebfa3087f3bfbe</td><td>Sparseness helps: Sparsity Augmented +<br/>Collaborative Representation for Classification +</td></tr><tr><td>d3b73e06d19da6b457924269bb208878160059da</td><td>Proceedings of the 5th International Conference on Computing and Informatics, ICOCI 2015 +<br/>11-13 August, 2015 Istanbul, Turkey. Universiti Utara Malaysia (http://www.uum.edu.my ) +<br/>Paper No. +<br/>065 +<br/>IMPLEMENTATION OF AN AUTOMATED SMART HOME +<br/>CONTROL FOR DETECTING HUMAN EMOTIONS VIA FACIAL +<br/>DETECTION +<br/>Osman4 +</td></tr><tr><td>d3d71a110f26872c69cf25df70043f7615edcf92</td><td>2736 +<br/>Learning Compact Feature Descriptor and Adaptive +<br/>Matching Framework for Face Recognition +<br/>improvements +</td></tr><tr><td>d309e414f0d6e56e7ba45736d28ee58ae2bad478</td><td>Efficient Two-Stream Motion and Appearance 3D CNNs for +<br/>Video Classification +<br/>Ali Diba +<br/>ESAT-KU Leuven +<br/>Ali Pazandeh +<br/>Sharif UTech +<br/>Luc Van Gool +<br/>ESAT-KU Leuven, ETH Zurich +</td></tr><tr><td>d3f5a1848b0028d8ab51d0b0673732cad2e3c8c9</td><td></td></tr><tr><td>d33fcdaf2c0bd0100ec94b2c437dccdacec66476</td><td>Neurons with Paraboloid Decision Boundaries for +<br/>Improved Neural Network Classification +<br/>Performance +</td></tr><tr><td>d444368421f456baf8c3cb089244e017f8d32c41</td><td>CNN for IMU Assisted Odometry Estimation using Velodyne LiDAR +</td></tr><tr><td>d4c7d1a7a03adb2338704d2be7467495f2eb6c7b</td><td></td></tr><tr><td>d4ebf0a4f48275ecd8dbc2840b2a31cc07bd676d</td><td></td></tr><tr><td>d44a93027208816b9e871101693b05adab576d89</td><td></td></tr><tr><td>d4b88be6ce77164f5eea1ed2b16b985c0670463a</td><td>TECHNICAL REPORT JAN.15.2016 +<br/>A Survey of Different 3D Face Reconstruction +<br/>Methods +<br/>Department of Computer Science and Engineering +</td></tr><tr><td>d44ca9e7690b88e813021e67b855d871cdb5022f</td><td>QUT Digital Repository: +<br/>http://eprints.qut.edu.au/ +<br/>Zhang, Ligang and Tjondronegoro, Dian W. (2009) Selecting, optimizing and +<br/>fusing ‘salient’ Gabor features for facial expression recognition. In: Neural +<br/>Information Processing (Lecture Notes in Computer Science), 1-5 December +<br/>2009, Hotel Windsor Suites Bangkok, Bangkok. +<br/> +<br/> © Copyright 2009 Springer-Verlag GmbH Berlin Heidelberg +<br/> +</td></tr><tr><td>bafb8812817db7445fe0e1362410a372578ec1fc</td><td>805 +<br/>Image-Quality-Based Adaptive Face Recognition +</td></tr><tr><td>ba816806adad2030e1939450226c8647105e101c</td><td>MindLAB at the THUMOS Challenge +<br/>Fabi´an P´aez +<br/>Fabio A. Gonz´alez +<br/>MindLAB Research Group +<br/>MindLAB Research Group +<br/>MindLAB Research Group +<br/>Bogot´a, Colombia +<br/>Bogot´a, Colombia +<br/>Bogot´a, Colombia +</td></tr><tr><td>badcd992266c6813063c153c41b87babc0ba36a3</td><td>Recent Advances in Object Detection in the Age +<br/>of Deep Convolutional Neural Networks +<br/>,1,2), Fr´ed´eric Jurie(1) +<br/>(∗) equal contribution +<br/>(1)Normandie Univ, UNICAEN, ENSICAEN, CNRS +<br/>(2)Safran Electronics and Defense +<br/>September 11, 2018 +</td></tr><tr><td>ba788365d70fa6c907b71a01d846532ba3110e31</td><td></td></tr><tr><td>ba8a99d35aee2c4e5e8a40abfdd37813bfdd0906</td><td>ELEKTROTEHNI ˇSKI VESTNIK 78(1-2): 12–17, 2011 +<br/>EXISTING SEPARATE ENGLISH EDITION +<br/>Uporaba emotivno pogojenega raˇcunalniˇstva v +<br/>priporoˇcilnih sistemih +<br/>Marko Tkalˇciˇc, Andrej Koˇsir, Jurij Tasiˇc +<br/>1Univerza v Ljubljani, Fakulteta za elektrotehniko, Trˇzaˇska 25, 1000 Ljubljana, Slovenija +<br/>2Univerza v Ljubljani, Fakulteta za raˇcunalniˇstvo in informatiko, Trˇzaˇska 25, 1000 Ljubljana, Slovenija +<br/>Povzetek. V ˇclanku predstavljamo rezultate treh raziskav, vezanih na izboljˇsanje delovanja multimedijskih +<br/>priporoˇcilnih sistemov s pomoˇcjo metod emotivno pogojenega raˇcunalniˇstva (ang. affective computing). +<br/>Vsebinski priporoˇcilni sistem smo izboljˇsali s pomoˇcjo metapodatkov, ki opisujejo emotivne odzive uporabnikov. +<br/>Pri skupinskem priporoˇcilnem sistemu smo dosegli znaˇcilno izboljˇsanje v obmoˇcju hladnega zagona z uvedbo +<br/>nove mere podobnosti, ki temelji na osebnostnem modelu velikih pet (ang. five factor model). Razvili smo tudi +<br/>sistem za neinvazivno oznaˇcevanje vsebin z emotivnimi parametri, ki pa ˇse ni zrel za uporabo v priporoˇcilnih +<br/>sistemih. +<br/>Kljuˇcne besede: priporoˇcilni sistemi, emotivno pogojeno raˇcunalniˇstvo, strojno uˇcenje, uporabniˇski profil, +<br/>emocije +<br/>Uporaba emotivnega raˇcunalniˇstva v priporoˇcilnih +<br/>sistemih +<br/>In this paper we present the results of three investigations of +<br/>our broad research on the usage of affect and personality in +<br/>recommender systems. We improved the accuracy of content- +<br/>based recommender system with the inclusion of affective +<br/>parameters of user and item modeling. We improved the +<br/>accuracy of a content filtering recommender system under the +<br/>cold start conditions with the introduction of a personality +<br/>based user similarity measure. Furthermore we developed a +<br/>system for implicit tagging of content with affective metadata. +<br/>1 UVOD +<br/>Uporabniki (porabniki) multimedijskih (MM) vsebin so +<br/>v ˇcedalje teˇzjem poloˇzaju, saj v veliki koliˇcini vse- +<br/>bin teˇzko najdejo zanje primerne. Pomagajo si s pri- +<br/>poroˇcilnimi sistemi, ki na podlagi osebnih preferenc +<br/>uporabnikov izberejo manjˇso koliˇcino relevantnih MM +<br/>vsebin, med katerimi uporabnik laˇze izbira. Noben danes +<br/>znan priporoˇcilni sistem ne zadoˇsˇca v celoti potrebam +<br/>uporabnikov, saj je izbor priporoˇcenih vsebin obiˇcajno +<br/>nezadovoljive kakovosti [10]. Cilj tega ˇclanka je pred- +<br/>staviti metode emotivno pogojenega raˇcunalniˇstva (ang. +<br/>affective computing - glej [12]) za izboljˇsanje kakovosti +<br/>priporoˇcilnih sistemov in utrditi za slovenski prostor +<br/>novo terminologijo. +<br/>1.1 Opis problema +<br/>Za izboljˇsanje kakovosti priporoˇcilnih sistemov sta +<br/>na voljo dve poti: (i) optimizacija algoritmov ali (ii) +<br/>uporaba boljˇsih znaˇcilk, ki bolje razloˇzijo neznano +<br/>Prejet 13. oktober, 2010 +<br/>Odobren 1. februar, 2011 +<br/>varianco [8]. V tem ˇclanku predstavljamo izboljˇsanje +<br/>priporoˇcilnih sistemov z uporabo novih znaˇcilk, ki te- +<br/>meljijo na emotivnih odzivih uporabnikov in na njiho- +<br/>vih osebnostnih lastnostih. Te znaˇcilke razloˇzijo velik +<br/>del uporabnikovih preferenc, ki se izraˇzajo v obliki +<br/>ocen posameznih vsebin (npr. Likertova lestvica, binarne +<br/>ocene itd.). Ocene vsebin se pri priporoˇcilnih sistemih +<br/>zajemajo eksplicitno (ocena) ali implicitno, pri ˇcemer o +<br/>oceni sklepamo na podlagi opazovanj (npr. ˇcas gledanja +<br/>kot indikator vˇseˇcnosti [7]. +<br/>Izboljˇsanja uˇcinkovitosti priporoˇcilnih sistemov smo +<br/>se lotili na treh podroˇcjih: (i) uporaba emotivnega +<br/>modeliranja uporabnikov v vsebinskem priporoˇcilnem +<br/>sistemu, (ii) neinvazivna (implicitna) detekcija emocij za +<br/>emotivno modeliranje in (iii) uporaba osebnostne mere +<br/>podobnosti v skupinskem priporoˇcilnem sistemu. Slika 1 +<br/>prikazuje arhitekturo emotivnega priporoˇcilnega sistema +<br/>in mesta, kjer smo vnesli opisane izboljˇsave. +<br/>Preostanek ˇclanka je strukturiran tako: v razdelku +<br/>2 je predstavljen zajem podatkov. V razdelku 3 je +<br/>predstavljen vsebinski priporoˇcilni sistem z emotivnimi +<br/>metapodatki. V razdelku 4 je predstavljen skupinski +<br/>priporoˇcilni sistem, ki uporablja mero podobnosti na +<br/>podlagi osebnosti, v razdelku 5 pa algoritem za razpo- +<br/>znavo emocij. Vsak od teh razdelov je sestavljen iz opisa +<br/>eksperimenta in predstavitve rezultatov. V razdelku 6 so +<br/>predstavljeni sklepi. +<br/>1.2 Sorodno delo +<br/>Najbolj groba delitev priporoˇcilnih sistemov je na vse- +<br/>binske, skupinske ter hibridne sisteme [1]. Z izjemo vse- +<br/>binskih priporoˇcilnih sistemov, ki sta ga razvila Arapakis +<br/>[2] in Tkalˇciˇc [14], sorodnega dela na podroˇcju emotivno +<br/>pogojenih priporoˇcilnih sistemov takorekoˇc ni. Panti´c in +</td></tr><tr><td>ba29ba8ec180690fca702ad5d516c3e43a7f0bb8</td><td></td></tr><tr><td>bab88235a30e179a6804f506004468aa8c28ce4f</td><td></td></tr><tr><td>badd371a49d2c4126df95120902a34f4bee01b00</td><td>GONDA, WEI, PARAG, PFISTER: PARALLEL SEPARABLE 3D CONVOLUTION +<br/>Parallel Separable 3D Convolution for Video +<br/>and Volumetric Data Understanding +<br/>Harvard John A. Paulson School of +<br/>Engineering and Applied Sciences +<br/>Camabridge MA, USA +<br/>Toufiq Parag +<br/>Hanspeter Pfister +</td></tr><tr><td>a0f94e9400938cbd05c4b60b06d9ed58c3458303</td><td>1118 +<br/>Value-Directed Human Behavior Analysis +<br/>from Video Using Partially Observable +<br/>Markov Decision Processes +</td></tr><tr><td>a022eff5470c3446aca683eae9c18319fd2406d5</td><td>2017-ENST-0071 +<br/>EDITE - ED 130 +<br/>Doctorat ParisTech +<br/>T H È S E +<br/>pour obtenir le grade de docteur délivré par +<br/>TÉLÉCOM ParisTech +<br/>Spécialité « SIGNAL et IMAGES » +<br/>présentée et soutenue publiquement par +<br/>le 15 décembre 2017 +<br/>Apprentissage Profond pour la Description Sémantique des Traits +<br/>Visuels Humains +<br/>Directeur de thèse : Jean-Luc DUGELAY +<br/>Co-encadrement de la thèse : Moez BACCOUCHE +<br/>Jury +<br/>Mme Bernadette DORIZZI, PRU, Télécom SudParis +<br/>Mme Jenny BENOIS-PINEAU, PRU, Université de Bordeaux +<br/>M. Christian WOLF, MC/HDR, INSA de Lyon +<br/>M. Patrick PEREZ, Chercheur/HDR, Technicolor Rennes +<br/>M. Moez BACCOUCHE, Chercheur/Docteur, Orange Labs Rennes +<br/>M. Jean-Luc DUGELAY, PRU, Eurecom Sophia Antipolis +<br/>M. Sid-Ahmed BERRANI, Directeur de l’Innovation/HDR, Algérie Télécom +<br/>Présidente +<br/>Rapporteur +<br/>Rapporteur +<br/>Examinateur +<br/>Encadrant +<br/>Directeur de Thèse +<br/>Invité +<br/>TÉLÉCOM ParisTech +<br/>école de l’Institut Télécom - membre de ParisTech +<br/>N°: 2009 ENAM XXXX T H È S E </td></tr><tr><td>a0c37f07710184597befaa7e6cf2f0893ff440e9</td><td></td></tr><tr><td>a0fb5b079dd1ee5ac6ac575fe29f4418fdb0e670</td><td></td></tr><tr><td>a0fd85b3400c7b3e11122f44dc5870ae2de9009a</td><td>Learning Deep Representation for Face +<br/>Alignment with Auxiliary Attributes +</td></tr><tr><td>a0dfb8aae58bd757b801e2dcb717a094013bc178</td><td>Reconocimiento de expresiones faciales con base +<br/>en la din´amica de puntos de referencia faciales +<br/>Instituto Nacional de Astrof´ısica ´Optica y Electr´onica, +<br/>Divisi´on de Ciencias Computacionales, Tonantzintla, Puebla, +<br/>M´exico +<br/>Resumen. Las expresiones faciales permiten a las personas comunicar +<br/>emociones, y es pr´acticamente lo primero que observamos al interactuar +<br/>con alguien. En el ´area de computaci´on, el reconocimiento de expresiones +<br/>faciales es importante debido a que su an´alisis tiene aplicaci´on directa en +<br/>´areas como psicolog´ıa, medicina, educaci´on, entre otras. En este articulo +<br/>se presenta el proceso de dise˜no de un sistema para el reconocimiento de +<br/>expresiones faciales utilizando la din´amica de puntos de referencia ubi- +<br/>cados en el rostro, su implementaci´on, experimentos realizados y algunos +<br/>de los resultados obtenidos hasta el momento. +<br/>Palabras clave: Expresiones faciales, clasificaci´on, m´aquinas de soporte +<br/>vectorial,modelos activos de apariencia. +<br/>Facial Expressions Recognition Based on Facial +<br/>Landmarks Dynamics +</td></tr><tr><td>a03cfd5c0059825c87d51f5dbf12f8a76fe9ff60</td><td>Simultaneous Learning and Alignment: +<br/>Multi-Instance and Multi-Pose Learning? +<br/>1 Comp. Science & Eng. +<br/>Univ. of CA, San Diego +<br/>2 Electrical Engineering +<br/>California Inst. of Tech. +<br/>3 Lab of Neuro Imaging +<br/>Univ. of CA, Los Angeles +</td></tr><tr><td>a000149e83b09d17e18ed9184155be140ae1266e</td><td>Chapter 9 +<br/>Action Recognition in Realistic +<br/>Sports Videos +</td></tr><tr><td>a784a0d1cea26f18626682ab108ce2c9221d1e53</td><td>Anchored Regression Networks applied to Age Estimation and Super Resolution +<br/>D-ITET, ETH Zurich +<br/>Switzerland +<br/>D-ITET, ETH Zurich +<br/>Merantix GmbH +<br/>D-ITET, ETH Zurich +<br/>ESAT, KU Leuven +</td></tr><tr><td>a74251efa970b92925b89eeef50a5e37d9281ad0</td><td></td></tr><tr><td>a7664247a37a89c74d0e1a1606a99119cffc41d4</td><td>Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence (IJCAI-17) +<br/>3287 +</td></tr><tr><td>a7a6eb53bee5e2224f2ecd56a14e3a5a717e55b9</td><td>11th International Symposium of Robotics Research (ISRR2003), pp.192-201, 2003 +<br/>Face Recognition Using Multi-viewpoint Patterns for +<br/>Robot Vision +<br/>Corporate Research and Development Center, TOSHIBA Corporation +<br/>1, KomukaiToshiba-cho, Saiwai-ku, Kawasaki 212-8582 Japan +</td></tr><tr><td>a75ee7f4c4130ef36d21582d5758f953dba03a01</td><td>DD2427 Final Project Report +<br/>DD2427 Final Project Report +<br/>Human face attributes prediction with Deep +<br/>Learning +</td></tr><tr><td>a775da3e6e6ea64bffab7f9baf665528644c7ed3</td><td>International Journal of Computer Applications (0975 – 8887) +<br/>Volume 142 – No.9, May 2016 +<br/>Human Face Pose Estimation based on Feature +<br/>Extraction Points +<br/>Research scholar, +<br/> Department of ECE +<br/>SBSSTC, Moga Road, +<br/> Ferozepur, Punjab, India +</td></tr><tr><td>b8dba0504d6b4b557d51a6cf4de5507141db60cf</td><td>Comparing Performances of Big Data Stream +<br/>Processing Platforms with RAM3S +</td></tr><tr><td>b8378ab83bc165bc0e3692f2ce593dcc713df34a</td><td></td></tr><tr><td>b8f3f6d8f188f65ca8ea2725b248397c7d1e662d</td><td>Selfie Detection by Synergy-Constriant Based +<br/>Convolutional Neural Network +<br/>Electrical and Electronics Engineering, NITK-Surathkal, India. +</td></tr><tr><td>b81cae2927598253da37954fb36a2549c5405cdb</td><td>Experiments on Visual Information Extraction with the Faces of Wikipedia +<br/>D´epartement de g´enie informatique et g´enie logiciel, Polytechnique Montr´eal +<br/>2500, Chemin de Polytechnique, Universit´e de Montr´eal, Montr`eal, Qu´ebec, Canada +</td></tr><tr><td>b8a829b30381106b806066d40dd372045d49178d</td><td>1872 +<br/>A Probabilistic Framework for Joint Pedestrian Head +<br/>and Body Orientation Estimation +</td></tr><tr><td>b1d89015f9b16515735d4140c84b0bacbbef19ac</td><td>Too Far to See? Not Really! +<br/>— Pedestrian Detection with Scale-aware +<br/>Localization Policy +</td></tr><tr><td>b14b672e09b5b2d984295dfafb05604492bfaec5</td><td>LearningImageClassificationandRetrievalModelsThomasMensink </td></tr><tr><td>b171f9e4245b52ff96790cf4f8d23e822c260780</td><td></td></tr><tr><td>b1a3b19700b8738b4510eecf78a35ff38406df22</td><td>This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TAFFC.2017.2731763, IEEE +<br/>Transactions on Affective Computing +<br/>JOURNAL OF LATEX CLASS FILES, VOL. 13, NO. 9, SEPTEMBER 2014 +<br/>Automatic Analysis of Facial Actions: A Survey +<br/>and Maja Pantic, Fellow, IEEE +</td></tr><tr><td>b1301c722886b6028d11e4c2084ee96466218be4</td><td></td></tr><tr><td>b1c5581f631dba78927aae4f86a839f43646220c</td><td></td></tr><tr><td>b1444b3bf15eec84f6d9a2ade7989bb980ea7bd1</td><td>LOCAL DIRECTIONAL RELATION PATTERN +<br/>Local Directional Relation Pattern for +<br/>Unconstrained and Robust Face Retrieval +</td></tr><tr><td>b19e83eda4a602abc5a8ef57467c5f47f493848d</td><td>JOURNAL OF LATEX CLASS FILES +<br/>Heat Kernel Based Local Binary Pattern for +<br/>Face Representation +</td></tr><tr><td>dd8084b2878ca95d8f14bae73e1072922f0cc5da</td><td>Model Distillation with Knowledge Transfer from +<br/>Face Classification to Alignment and Verification +<br/>Beijing Orion Star Technology Co., Ltd. Beijing, China +</td></tr><tr><td>dd0760bda44d4e222c0a54d41681f97b3270122b</td><td></td></tr><tr><td>ddea3c352f5041fb34433b635399711a90fde0e8</td><td>Facial Expression Classification using Visual Cues and Language +<br/>Department of Computer Science and Engineering, IIT Kanpur +</td></tr><tr><td>ddbd24a73ba3d74028596f393bb07a6b87a469c0</td><td>Multi-region two-stream R-CNN +<br/>for action detection +<br/>Inria(cid:63) +</td></tr><tr><td>ddf099f0e0631da4a6396a17829160301796151c</td><td>IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY +<br/>Learning Face Image Quality from +<br/>Human Assessments +</td></tr><tr><td>dd0a334b767e0065c730873a95312a89ef7d1c03</td><td>Eigenexpressions: Emotion Recognition using Multiple +<br/>Eigenspaces +<br/>Luis Marco-Gim´enez1, Miguel Arevalillo-Herr´aez1, and Cristina Cuhna-P´erez2 +<br/><b></b><br/>Burjassot. Valencia 46100, Spain, +<br/>2 Universidad Cat´olica San Vicente M´artir de Valencia (UCV), +<br/>Burjassot. Valencia. Spain +</td></tr><tr><td>dd2f6a1ba3650075245a422319d86002e1e87808</td><td></td></tr><tr><td>dd8d53e67668067fd290eb500d7dfab5b6f730dd</td><td>69 +<br/>A Parameter-Free Framework for General +<br/>Supervised Subspace Learning +</td></tr><tr><td>ddbb6e0913ac127004be73e2d4097513a8f02d37</td><td>264 +<br/>IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 1, NO. 3, SEPTEMBER 1999 +<br/>Face Detection Using Quantized Skin Color +<br/>Regions Merging and Wavelet Packet Analysis +</td></tr><tr><td>dd600e7d6e4443ebe87ab864d62e2f4316431293</td><td></td></tr><tr><td>dcb44fc19c1949b1eda9abe998935d567498467d</td><td>Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence (IJCAI-17) +<br/>1916 +</td></tr><tr><td>dc77287bb1fcf64358767dc5b5a8a79ed9abaa53</td><td>Fashion Conversation Data on Instagram +<br/>∗Graduate School of Culture Technology, KAIST, South Korea +<br/>†Department of Communication Studies, UCLA, USA +</td></tr><tr><td>dc2e805d0038f9d1b3d1bc79192f1d90f6091ecb</td><td></td></tr><tr><td>dc974c31201b6da32f48ef81ae5a9042512705fe</td><td>Am I done? Predicting Action Progress in Video +<br/>1 Media Integration and Communication Center, Univ. of Florence, Italy +<br/>2 Department of Mathematics “Tullio Levi-Civita”, Univ. of Padova, Italy +</td></tr><tr><td>b6c047ab10dd86b1443b088029ffe05d79bbe257</td><td></td></tr><tr><td>b6c53891dff24caa1f2e690552a1a5921554f994</td><td></td></tr><tr><td>b613b30a7cbe76700855479a8d25164fa7b6b9f1</td><td>1 +<br/>Identifying User-Specific Facial Affects from +<br/>Spontaneous Expressions with Minimal Annotation +</td></tr><tr><td>b6f682648418422e992e3ef78a6965773550d36b</td><td>February 8, 2017 +</td></tr><tr><td>b656abc4d1e9c8dc699906b70d6fcd609fae8182</td><td></td></tr><tr><td>a9eb6e436cfcbded5a9f4b82f6b914c7f390adbd</td><td>(IJARAI) International Journal of Advanced Research in Artificial Intelligence, +<br/>Vol. 5, No.6, 2016 +<br/>A Model for Facial Emotion Inference Based on +<br/>Planar Dynamic Emotional Surfaces +<br/>Ruivo, J. P. P. +<br/>Escola Polit´ecnica +<br/>Negreiros, T. +<br/>Escola Polit´ecnica +<br/>Barretto, M. R. P. +<br/>Escola Polit´ecnica +<br/>Tinen, B. +<br/>Escola Polit´ecnica +<br/>Universidade de S˜ao Paulo +<br/>Universidade de S˜ao Paulo +<br/>Universidade de S˜ao Paulo +<br/>Universidade de S˜ao Paulo +<br/>S˜ao Paulo, Brazil +<br/>S˜ao Paulo, Brazil +<br/>S˜ao Paulo, Brazil +<br/>S˜ao Paulo, Brazil +</td></tr><tr><td>a92adfdd8996ab2bd7cdc910ea1d3db03c66d34f</td><td></td></tr><tr><td>a98316980b126f90514f33214dde51813693fe0d</td><td>Collaborations on YouTube: From Unsupervised Detection to the +<br/>Impact on Video and Channel Popularity +<br/>Multimedia Communications Lab (KOM), Technische Universität Darmstadt, Germany +</td></tr><tr><td>a93781e6db8c03668f277676d901905ef44ae49f</td><td>Recent Datasets on Object Manipulation: A Survey +</td></tr><tr><td>a9adb6dcccab2d45828e11a6f152530ba8066de6</td><td>Aydınlanma Alt-uzaylarına dayalı Gürbüz Yüz Tanıma +<br/>Illumination Subspaces based Robust Face Recognition +<br/>Interactive Systems Labs, Universität Karlsruhe (TH) +<br/>76131 Karlsruhe, Almanya +<br/>web: http://isl.ira.uka.de/face_recognition +<br/>Özetçe +<br/>yönlerine +<br/>aydınlanma +<br/>kaynaklanan +<br/>sonra, yüz uzayı +<br/>Bu çalışmada aydınlanma alt-uzaylarına dayalı bir yüz tanıma +<br/>sistemi sunulmuştur. Bu sistemde, +<br/>ilk olarak, baskın +<br/>aydınlanma yönleri, bir topaklandırma algoritması kullanılarak +<br/>öğrenilmiştir. Topaklandırma algoritması sonucu önden, sağ +<br/>ve sol yanlardan olmak üzere üç baskın aydınlanma yönü +<br/>gözlemlenmiştir. Baskın +<br/>karar +<br/>-yüzün görünümündeki +<br/>kılındıktan +<br/>aydınlanmadan +<br/>kişi +<br/>kimliklerinden kaynaklanan değişimlerden ayırmak için- bu üç +<br/>aydınlanma uzayına bölünmüştür. Daha sonra, ek aydınlanma +<br/>yönü bilgisinden faydalanmak için aydınlanma alt-uzaylarına +<br/>dayalı yüz +<br/>tanıma algoritması kullanılmıştır. Önerilen +<br/>yaklaşım, CMU PIE veritabanında, “illumination” ve +<br/>“lighting” kümelerinde yer alan yüz +<br/>imgeleri üzerinde +<br/>sınanmıştır. Elde edilen deneysel sonuçlar, aydınlanma +<br/>yönünden yararlanmanın ve aydınlanma alt-uzaylarına dayalı +<br/>yüz tanıma algoritmasının yüz tanıma başarımını önemli +<br/>ölçüde arttırdığını göstermiştir. +<br/>değişimleri, +<br/>farklı +</td></tr><tr><td>a95dc0c4a9d882a903ce8c70e80399f38d2dcc89</td><td> TR-IIS-14-003 +<br/>Review and Implementation of +<br/>High-Dimensional Local Binary +<br/>Patterns and Its Application to +<br/>Face Recognition +<br/>July. 24, 2014 || Technical Report No. TR-IIS-14-003 +<br/>http://www.iis.sinica.edu.tw/page/library/TechReport/tr2014/tr14.html +</td></tr><tr><td>a9286519e12675302b1d7d2fe0ca3cc4dc7d17f6</td><td>Learning to Succeed while Teaching to Fail: +<br/>Privacy in Closed Machine Learning Systems +</td></tr><tr><td>a92b5234b8b73e06709dd48ec5f0ec357c1aabed</td><td></td></tr><tr><td>d50c6d22449cc9170ab868b42f8c72f8d31f9b6c</td><td>Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence (IJCAI-17) +<br/>1668 +</td></tr><tr><td>d522c162bd03e935b1417f2e564d1357e98826d2</td><td>He et al. EURASIP Journal on Advances in Signal Processing 2013, 2013:19 +<br/>http://asp.eurasipjournals.com/content/2013/1/19 +<br/>RESEARCH +<br/>Open Access +<br/>Weakly supervised object extraction with +<br/>iterative contour prior for remote sensing +<br/>images +</td></tr><tr><td>d59f18fcb07648381aa5232842eabba1db52383e</td><td>International Conference on Systemics, Cybernetics and Informatics, February 12–15, 2004 +<br/>ROBUST FACIAL EXPRESSION RECOGNITION USING SPATIALLY +<br/>LOCALIZED GEOMETRIC MODEL +<br/>Department of Electrical Engineering +<br/>Dept. of Computer Sc. and Engg. +<br/>IIT Kanpur +<br/> Kanpur 208016, India +<br/>Kanpur 208016, India +<br/> IIT Kanpur +<br/>Dept. of Computer Sc. and Engg. +<br/> IIT Kanpur +<br/>Kanpur 208016, India +<br/>While approaches based on 3D deformable facial model have +<br/>achieved expression recognition rates of as high as 98% [2], they +<br/>are computationally inefficient and require considerable apriori +<br/>training based on 3D information, which is often unavailable. +<br/>Recognition from 2D images remains a difficult yet important +<br/>problem for areas such as +<br/>image database querying and +<br/>classification. The accuracy rates achieved for 2D images are +<br/>around 90% [3,4,5,11]. In a recent review of expression +<br/>recognition, Fasel [1] considers the problem along several +<br/>dimensions: whether features such as lips or eyebrows are first +<br/>identified in the face (local [4] vs holistic [11]), or whether the +<br/>image model used is 2D or 3D. Methods proposed for expression +<br/>recognition from 2D images include the Gabor-Wavelet [5] or +<br/>Holistic Optical flow [11] approach. +<br/>This paper describes a more robust system for facial expression +<br/>recognition from image sequences using 2D appearance-based +<br/>local approach for the extraction of intransient facial features, i.e. +<br/>features such as eyebrows, lips, or mouth, which are always +<br/>present in the image, but may be deformed [1] (in contrast, +<br/>transient features are wrinkles or bulges that disappear at other +<br/>times). The main advantages of such an approach is low +<br/>computational requirements, ability to work with both colored and +<br/>grayscale images and robustness in handling partial occlusions +<br/>[3]. +<br/>Edge projection analysis which is used here for feature extraction +<br/>(eyebrows and lips) is well known [6]. Unlike [6] which describes +<br/>a template based matching as an essential starting point, we use +<br/>contours analysis. Our system computes a feature vector based on +<br/>geometrical model of the face and then classifies it into four +<br/>expression classes using a feed-forward basis function net. The +<br/>system detects open and closed state of the mouth as well. The +<br/>algorithm presented here works on both color and grayscale image +<br/>sequences. An important aspect of our work is the use of color +<br/>information for robust and more accurate segmentation of lip +<br/>region in case of color images. The novel lip-enhancement +<br/>transform is based on Gaussian modeling of skin and lip color. +<br/>To place the work in a larger context of face analysis and +<br/>recognition, the overall task requires that the part of the image +<br/>involving the face be detected and segmented. We assume that a +<br/>near-frontal view of the face is available. Tests on a grayscale +<br/>and two color face image databases ([8] and [9,10]) demonstrate a +<br/>superior recognition rate for four facial expressions (smile, +<br/>surprise, disgust and sad against neutral). +<br/>image sequences +</td></tr><tr><td>d588dd4f305cdea37add2e9bb3d769df98efe880</td><td> +<br/>Audio-Visual Authentication System over the +<br/>Internet Protocol +<br/>abandoned. +<br/>in +<br/>illumination based +<br/>is developed with the objective to +</td></tr><tr><td>d5444f9475253bbcfef85c351ea9dab56793b9ea</td><td>IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS +<br/>BoxCars: Improving Fine-Grained Recognition +<br/>of Vehicles using 3D Bounding Boxes +<br/>in Traffic Surveillance +<br/>in contrast +</td></tr><tr><td>d5ab6aa15dad26a6ace5ab83ce62b7467a18a88e</td><td>World Journal of Computer Application and Technology 2(7): 133-138, 2014 +<br/>DOI: 10.13189/wjcat.2014.020701 +<br/>http://www.hrpub.org +<br/>Optimized Structure for Facial Action Unit Relationship +<br/>Using Bayesian Network +<br/>Intelligent Biometric Group, School of Electrical and Electronic Engineering, Engineering Campus, Universiti Sains Malaysia, Pulau +<br/>Pinang, Malaysia +<br/>Copyright © 2014 Horizon Research Publishing All rights reserved. +</td></tr><tr><td>d56fe69cbfd08525f20679ffc50707b738b88031</td><td>Training of multiple classifier systems utilizing +<br/>partially labelled sequences +<br/><b></b><br/>89069 Ulm - Germany +</td></tr><tr><td>d50751da2997e7ebc89244c88a4d0d18405e8507</td><td></td></tr><tr><td>d511e903a882658c9f6f930d6dd183007f508eda</td><td></td></tr><tr><td>d59404354f84ad98fa809fd1295608bf3d658bdc</td><td>International Journal of Computer Vision manuscript No. +<br/>(will be inserted by the editor) +<br/>Face Synthesis from Visual Attributes via Sketch using +<br/>Conditional VAEs and GANs +<br/>Received: date / Accepted: date +</td></tr><tr><td>d5e1173dcb2a51b483f86694889b015d55094634</td><td></td></tr><tr><td>d2eb1079552fb736e3ba5e494543e67620832c52</td><td>ANNUNZIATA, SAGONAS, CALÌ: DENSELY FUSED SPATIAL TRANSFORMER NETWORKS1 +<br/>DeSTNet: Densely Fused Spatial +<br/>Transformer Networks1 +<br/>Onfido Research +<br/>3 Finsbury Avenue +<br/>London, UK +</td></tr><tr><td>d24dafe10ec43ac8fb98715b0e0bd8e479985260</td><td>J Nonverbal Behav (2018) 42:81–99 +<br/>https://doi.org/10.1007/s10919-017-0266-z +<br/>O R I G I N A L P A P E R +<br/>Effects of Social Anxiety on Emotional Mimicry +<br/>and Contagion: Feeling Negative, but Smiling Politely +<br/>• Gerben A. van Kleef2 +<br/>• Agneta H. Fischer2 +<br/>Published online: 25 September 2017 +<br/>Ó The Author(s) 2017. This article is an open access publication +</td></tr><tr><td>d278e020be85a1ccd90aa366b70c43884dd3f798</td><td>Learning From Less Data: Diversified Subset Selection and +<br/>Active Learning in Image Classification Tasks +<br/>IIT Bombay +<br/>Mumbai, Maharashtra, India +<br/>AITOE Labs +<br/>Mumbai, Maharashtra, India +<br/>AITOE Labs +<br/>Mumbai, Maharashtra, India +<br/>Rishabh Iyer +<br/>AITOE Labs +<br/>Seattle, Washington, USA +<br/>AITOE Labs +<br/>Seattle, Washington, USA +<br/>Narsimha Raju +<br/>IIT Bombay +<br/>Mumbai, Maharashtra, India +<br/>IIT Bombay +<br/>Mumbai, Maharashtra, India +<br/>IIT Bombay +<br/>Mumbai, Maharashtra, India +<br/>May 30, 2018 +</td></tr><tr><td>aafb271684a52a0b23debb3a5793eb618940c5dd</td><td></td></tr><tr><td>aa52910c8f95e91e9fc96a1aefd406ffa66d797d</td><td>FACE RECOGNITION SYSTEM BASED +<br/>ON 2DFLD AND PCA +<br/>E&TC Department +<br/>Sinhgad Academy of Engineering +<br/>Pune, India +<br/>Mr. Hulle Rohit Rajiv +<br/>ME E&TC [Digital System] +<br/>Sinhgad Academy of Engineering +<br/>Pune, India +</td></tr><tr><td>aadfcaf601630bdc2af11c00eb34220da59b7559</td><td>Multi-view Hybrid Embedding: +<br/>A Divide-and-Conquer Approach +</td></tr><tr><td>aaa4c625f5f9b65c7f3df5c7bfe8a6595d0195a5</td><td>Biometrics in Ambient Intelligence +</td></tr><tr><td>aa331fe378056b6d6031bb8fe6676e035ed60d6d</td><td></td></tr><tr><td>aae0e417bbfba701a1183d3d92cc7ad550ee59c3</td><td>844 +<br/>A Statistical Method for 2-D Facial Landmarking +</td></tr><tr><td>aa577652ce4dad3ca3dde44f881972ae6e1acce7</td><td>Deep Attribute Networks +<br/>Department of EE, KAIST +<br/>Daejeon, South Korea +<br/>Department of EE, KAIST +<br/>Daejeon, South Korea +<br/>Department of EE, KAIST +<br/>Daejeon, South Korea +<br/>Department of EE, KAIST +<br/>Daejeon, South Korea +</td></tr><tr><td>aa94f214bb3e14842e4056fdef834a51aecef39c</td><td>Reconhecimento de padrões faciais: Um estudo +<br/>Universidade Federal +<br/>Rural do Semi-Árido +<br/>Departamento de Ciências Naturais +<br/>Mossoró, RN - 59625-900 +<br/>Resumo—O reconhecimento facial tem sido utilizado em di- +<br/>versas áreas para identificação e autenticação de usuários. Um +<br/>dos principais mercados está relacionado a segurança, porém há +<br/>uma grande variedade de aplicações relacionadas ao uso pessoal, +<br/>conveniência, aumento de produtividade, etc. O rosto humano +<br/>possui um conjunto de padrões complexos e mutáveis. Para +<br/>reconhecer esses padrões, são necessárias técnicas avançadas de +<br/>reconhecimento de padrões capazes, não apenas de reconhecer, +<br/>mas de se adaptar às mudanças constantes das faces das pessoas. +<br/>Este documento apresenta um método de reconhecimento facial +<br/>proposto a partir da análise comparativa de trabalhos encontra- +<br/>dos na literatura. +<br/>biométrica é o uso da biometria para reconhecimento, identi- +<br/>ficação ou verificação, de um ou mais traços biométricos de +<br/>um indivíduo com o objetivo de autenticar sua identidade. Os +<br/>traços biométricos são os atributos analisados pelas técnicas +<br/>de reconhecimento biométrico. +<br/>A tarefa de reconhecimento facial é composta por três +<br/>processos distintos: Registro, verificação e identificação bio- +<br/>métrica. Os processos se diferenciam pela forma de determinar +<br/>a identidade de um indivíduo. Na Figura 1 são descritos os +<br/>processos de registro, verificação e identificação biométrica. +<br/>I. INTRODUÇÃO +<br/>Biometria é a ciência que estabelece a identidade de um +<br/>indivíduo baseada em seus atributos físicos, químicos ou +<br/>comportamentais [1]. Possui inúmeras aplicações em diver- +<br/>sas áreas, se destacando mais na área de segurança, como +<br/>por exemplo sistemas de gerenciamento de identidade, cuja +<br/>funcionalidade é autenticar a identidade de um indivíduo no +<br/>contexto de uma aplicação. +<br/>O reconhecimento facial é uma técnica biométrica que +<br/>consiste em identificar padrões em características faciais como +<br/>formato da boca, do rosto, distância dos olhos, entre outros. +<br/>Um humano é capaz de reconhecer uma pessoa familiar +<br/>mesmo com muitos obstáculos com distância, sombras ou +<br/>apenas a visão parcial do rosto. Uma máquina, no entanto, +<br/>precisa realizar inúmeros processos para detectar e reconhecer +<br/>um conjunto de padrões específicos para rotular uma face +<br/>como conhecida ou desconhecida. Para isso, exitem métodos +<br/>capazes de detectar, extrair e classificar as características +<br/>faciais, fornecendo um reconhecimento automático de pessoas. +<br/>II. RECONHECIMENTO FACIAL +<br/>A tecnologia biométrica oferece vantagens em relação a +<br/>outros métodos tradicionais de identificação como senhas, +<br/>documentos e tokens. Entre elas estão o fato de que os +<br/>traços biométricos não podem ser perdidos ou esquecidos, são +<br/>difíceis de serem copiados, compartilhados ou distribuídos. Os +<br/>métodos requerem que a pessoa autenticada esteja presente +<br/>na hora e lugar da autenticação, evitando que pessoas má +<br/>intencionadas tenham acesso sem autorização. +<br/>A autenticação é o ato de estabelecer ou confirmar alguém, +<br/>ou alguma coisa, como autêntico, isto é, que as alegações +<br/>feitas por ou sobre a coisa é verdadeira [2]. Autenticação +<br/>(a) +<br/>(b) +<br/>(c) +<br/>Figura 1: Registro biométrico (a), identificação biométrica (b) +<br/>e verificação biométrica (c) +<br/>A Figura 1a descreve o processo de registro de dados +</td></tr><tr><td>af8fe1b602452cf7fc9ecea0fd4508ed4149834e</td><td></td></tr><tr><td>af6e351d58dba0962d6eb1baf4c9a776eb73533f</td><td>How to Train Your Deep Neural Network with +<br/>Dictionary Learning +<br/>*IIIT Delhi +<br/>Okhla Phase 3 +<br/>Delhi, 110020, India +<br/>+IIIT Delhi +<br/>Okhla Phase 3 +<br/>#IIIT Delhi +<br/>Okhla Phase 3 +<br/>Delhi, 110020, India +<br/>Delhi, 110020, India +</td></tr><tr><td>af6cae71f24ea8f457e581bfe1240d5fa63faaf7</td><td></td></tr><tr><td>af54dd5da722e104740f9b6f261df9d4688a9712</td><td></td></tr><tr><td>afc7092987f0d05f5685e9332d83c4b27612f964</td><td>Person-Independent Facial Expression Detection using Constrained +<br/>Local Models +</td></tr><tr><td>b730908bc1f80b711c031f3ea459e4de09a3d324</td><td>2024 +<br/>Active Orientation Models for Face +<br/>Alignment In-the-Wild +</td></tr><tr><td>b7cf7bb574b2369f4d7ebc3866b461634147041a</td><td>Neural Comput & Applic (2012) 21:1575–1583 +<br/>DOI 10.1007/s00521-011-0728-x +<br/>O R I G I N A L A R T I C L E +<br/>From NLDA to LDA/GSVD: a modified NLDA algorithm +<br/>Received: 2 August 2010 / Accepted: 3 August 2011 / Published online: 19 August 2011 +<br/>Ó Springer-Verlag London Limited 2011 +</td></tr><tr><td>b7eead8586ffe069edd190956bd338d82c69f880</td><td>A VIDEO DATABASE FOR FACIAL +<br/>BEHAVIOR UNDERSTANDING +<br/>D. Freire-Obreg´on and M. Castrill´on-Santana. +<br/>SIANI, Universidad de Las Palmas de Gran Canaria, Spain +</td></tr><tr><td>b75cee96293c11fe77ab733fc1147950abbe16f9</td><td></td></tr><tr><td>b7f05d0771da64192f73bdb2535925b0e238d233</td><td> MVA2005 IAPR Conference on Machine VIsion Applications, May 16-18, 2005 Tsukuba Science City, Japan +<br/>4-3 +<br/>Robust Active Shape Model using AdaBoosted Histogram Classifiers +<br/>W ataru Ito +<br/>Imaging Software Technology Center +<br/>Imaging Software Technology Center +<br/>FUJI PHOTO FILM CO., LTD. +<br/>FUJI PHOTO FILM CO., LTD. +</td></tr><tr><td>b755505bdd5af078e06427d34b6ac2530ba69b12</td><td>To appear in the International Joint Conf. Biometrics, Washington D.C., October, 2011 +<br/>NFRAD: Near-Infrared Face Recognition at a Distance +<br/>aDept. of Brain and Cognitive Eng. Korea Univ., Seoul, Korea +<br/>bDept. of Comp. Sci. & Eng. Michigan State Univ., E. Lansing, MI, USA 48824 +</td></tr><tr><td>b73fdae232270404f96754329a1a18768974d3f6</td><td></td></tr><tr><td>b76af8fcf9a3ebc421b075b689defb6dc4282670</td><td>Face Mask Extraction in Video Sequence +</td></tr><tr><td>b747fcad32484dfbe29530a15776d0df5688a7db</td><td></td></tr><tr><td>b7f7a4df251ff26aca83d66d6b479f1dc6cd1085</td><td>Bouges et al. EURASIP Journal on Image and Video Processing 2013, 2013:55 +<br/>http://jivp.eurasipjournals.com/content/2013/1/55 +<br/>RESEARCH +<br/>Open Access +<br/>Handling missing weak classifiers in boosted +<br/>cascade: application to multiview and +<br/>occluded face detection +</td></tr><tr><td>db227f72bb13a5acca549fab0dc76bce1fb3b948</td><td>International Refereed Journal of Engineering and Science (IRJES) +<br/>ISSN (Online) 2319-183X, (Print) 2319-1821 +<br/>Volume 4, Issue 6 (June 2015), PP.169-169-174 +<br/>Characteristic Based Image Search using Re-Ranking method +<br/>1Chitti Babu, 2Yasmeen Jaweed, 3G.Vijay Kumar +<br/><b></b></td></tr><tr><td>dbaf89ca98dda2c99157c46abd136ace5bdc33b3</td><td>Nonlinear Cross-View Sample Enrichment for +<br/>Action Recognition +<br/>Institut Mines-T´el´ecom; T´el´ecom ParisTech; CNRS LTCI +</td></tr><tr><td>dbab6ac1a9516c360cdbfd5f3239a351a64adde7</td><td></td></tr><tr><td>dbe255d3d2a5d960daaaba71cb0da292e0af36a7</td><td>Evolutionary Cost-sensitive Extreme Learning +<br/>Machine +<br/>1 +</td></tr><tr><td>dbb0a527612c828d43bcb9a9c41f1bf7110b1dc8</td><td>Chapter 7 +<br/>Machine Learning Techniques +<br/>for Face Analysis +</td></tr><tr><td>dbb7f37fb9b41d1aa862aaf2d2e721a470fd2c57</td><td>Face Image Analysis With +<br/>Convolutional Neural Networks +<br/>Dissertation +<br/>Zur Erlangung des Doktorgrades +<br/>der Fakult¨at f¨ur Angewandte Wissenschaften +<br/>an der Albert-Ludwigs-Universit¨at Freiburg im Breisgau +<br/>von +<br/>Stefan Duffner +<br/>2007 +</td></tr><tr><td>a83fc450c124b7e640adc762e95e3bb6b423b310</td><td>Deep Face Feature for Face Alignment +</td></tr><tr><td>a85e9e11db5665c89b057a124547377d3e1c27ef</td><td>Dynamics of Driver’s Gaze: Explorations in +<br/>Behavior Modeling & Maneuver Prediction +</td></tr><tr><td>a8117a4733cce9148c35fb6888962f665ae65b1e</td><td>IEEE TRANSACTIONS ON XXXX, VOL. XX, NO. XX, XX 201X +<br/>A Good Practice Towards Top Performance of Face +<br/>Recognition: Transferred Deep Feature Fusion +</td></tr><tr><td>a8035ca71af8cc68b3e0ac9190a89fed50c92332</td><td>000 +<br/>001 +<br/>002 +<br/>003 +<br/>004 +<br/>005 +<br/>006 +<br/>007 +<br/>008 +<br/>009 +<br/>010 +<br/>011 +<br/>012 +<br/>013 +<br/>014 +<br/>015 +<br/>016 +<br/>017 +<br/>018 +<br/>019 +<br/>020 +<br/>021 +<br/>022 +<br/>023 +<br/>024 +<br/>025 +<br/>026 +<br/>027 +<br/>028 +<br/>029 +<br/>030 +<br/>031 +<br/>032 +<br/>033 +<br/>034 +<br/>035 +<br/>036 +<br/>037 +<br/>038 +<br/>039 +<br/>040 +<br/>041 +<br/>042 +<br/>043 +<br/>044 +<br/>IIIT-CFW: A Benchmark Database of +<br/>Cartoon Faces in the Wild +<br/>1 IIIT Chittoor, Sri City, India +<br/>2 CVIT, KCIS, IIIT Hyderabad, India +</td></tr><tr><td>a88640045d13fc0207ac816b0bb532e42bcccf36</td><td>ARXIV VERSION +<br/>Simultaneously Learning Neighborship and +<br/>Projection Matrix for Supervised +<br/>Dimensionality Reduction +</td></tr><tr><td>a8a30a8c50d9c4bb8e6d2dd84bc5b8b7f2c84dd8</td><td>This is a repository copy of Modelling of Orthogonal Craniofacial Profiles. +<br/>White Rose Research Online URL for this paper: +<br/>http://eprints.whiterose.ac.uk/131767/ +<br/>Version: Published Version +<br/>Article: +<br/>Dai, Hang, Pears, Nicholas Edwin orcid.org/0000-0001-9513-5634 and Duncan, Christian +<br/>(2017) Modelling of Orthogonal Craniofacial Profiles. Journal of Imaging. ISSN 2313-433X +<br/>https://doi.org/10.3390/jimaging3040055 +<br/>Reuse +<br/>This article is distributed under the terms of the Creative Commons Attribution (CC BY) licence. This licence +<br/>allows you to distribute, remix, tweak, and build upon the work, even commercially, as long as you credit the +<br/>authors for the original work. More information and the full terms of the licence here: +<br/>https://creativecommons.org/licenses/ +<br/>Takedown +<br/>If you consider content in White Rose Research Online to be in breach of UK law, please notify us by +<br/>https://eprints.whiterose.ac.uk/ +</td></tr><tr><td>a8e75978a5335fd3deb04572bb6ca43dbfad4738</td><td>Sparse Graphical Representation based Discriminant +<br/>Analysis for Heterogeneous Face Recognition +</td></tr><tr><td>ded968b97bd59465d5ccda4f1e441f24bac7ede5</td><td>Noname manuscript No. +<br/>(will be inserted by the editor) +<br/>Large scale 3D Morphable Models +<br/>Zafeiriou +<br/>Received: date / Accepted: date +</td></tr><tr><td>de0eb358b890d92e8f67592c6e23f0e3b2ba3f66</td><td>ACCEPTED BY IEEE TRANS. PATTERN ANAL. AND MACH. INTELL. +<br/>Inference-Based Similarity Search in +<br/>Randomized Montgomery Domains for +<br/>Privacy-Preserving Biometric Identification +</td></tr><tr><td>dee406a7aaa0f4c9d64b7550e633d81bc66ff451</td><td>Content-Adaptive Sketch Portrait Generation by +<br/>Decompositional Representation Learning +</td></tr><tr><td>dedabf9afe2ae4a1ace1279150e5f1d495e565da</td><td>3294 +<br/>Robust Face Recognition With Structurally +<br/>Incoherent Low-Rank Matrix Decomposition +</td></tr><tr><td>de398bd8b7b57a3362c0c677ba8bf9f1d8ade583</td><td>Hierarchical Bayesian Theme Models for +<br/>Multi-pose Facial Expression Recognition +</td></tr><tr><td>ded41c9b027c8a7f4800e61b7cfb793edaeb2817</td><td></td></tr><tr><td>defa8774d3c6ad46d4db4959d8510b44751361d8</td><td>FEBEI - Face Expression Based Emoticon Identification +<br/>CS - B657 Computer Vision +<br/>Robert J Henderson - rojahend +</td></tr><tr><td>b0c512fcfb7bd6c500429cbda963e28850f2e948</td><td></td></tr><tr><td>b09b693708f412823053508578df289b8403100a</td><td>WANG et al.: TWO-STREAM SR-CNNS FOR ACTION RECOGNITION IN VIDEOS +<br/>Two-Stream SR-CNNs for Action +<br/>Recognition in Videos +<br/>1 Advanced Interactive Technologies Lab +<br/>ETH Zurich +<br/>Zurich, Switzerland +<br/>2 Computer Vision Lab +<br/>ETH Zurich +<br/>Zurich, Switzerland +</td></tr><tr><td>b07582d1a59a9c6f029d0d8328414c7bef64dca0</td><td>Employing Fusion of Learned and Handcrafted +<br/>Features for Unconstrained Ear Recognition +<br/>Maur´ıcio Pamplona Segundo∗† +<br/>October 24, 2017 +</td></tr><tr><td>b03d6e268cde7380e090ddaea889c75f64560891</td><td></td></tr><tr><td>b0c1615ebcad516b5a26d45be58068673e2ff217</td><td>How Image Degradations Affect Deep CNN-based Face +<br/>Recognition? +<br/>S¸amil Karahan1 Merve Kılınc¸ Yıldırım1 Kadir Kırtac¸1 Ferhat S¸ ¨ukr¨u Rende1 +<br/>G¨ultekin B¨ut¨un1Hazım Kemal Ekenel2 +</td></tr><tr><td>b0de0892d2092c8c70aa22500fed31aa7eb4dd3f</td><td>(will be inserted by the editor) +<br/>A robust and efficient video representation for action recognition +<br/>Received: date / Accepted: date +</td></tr><tr><td>a66d89357ada66d98d242c124e1e8d96ac9b37a0</td><td>Failure Detection for Facial Landmark Detectors +<br/>Computer Vision Lab, D-ITET, ETH Zurich, Switzerland +</td></tr><tr><td>a608c5f8fd42af6e9bd332ab516c8c2af7063c61</td><td>2408 +<br/>Age Estimation via Grouping and Decision Fusion +</td></tr><tr><td>a6eb6ad9142130406fb4ffd4d60e8348c2442c29</td><td>Video Description: A Survey of Methods, +<br/>Datasets and Evaluation Metrics +</td></tr><tr><td>a6583c8daa7927eedb3e892a60fc88bdfe89a486</td><td></td></tr><tr><td>a6590c49e44aa4975b2b0152ee21ac8af3097d80</td><td>https://doi.org/10.1007/s11263-018-1074-6 +<br/>3D Interpreter Networks for Viewer-Centered Wireframe Modeling +<br/>Received: date / Accepted: date +</td></tr><tr><td>a694180a683f7f4361042c61648aa97d222602db</td><td>Face Recognition using Scattering Wavelet under Illicit Drug Abuse Variations +<br/>IIIT-Delhi India +</td></tr><tr><td>a6db73f10084ce6a4186363ea9d7475a9a658a11</td><td></td></tr><tr><td>a6634ff2f9c480e94ed8c01d64c9eb70e0d98487</td><td></td></tr><tr><td>b9d0774b0321a5cfc75471b62c8c5ef6c15527f5</td><td>Fishy Faces: Crafting Adversarial Images to Poison Face Authentication +<br/>imec-DistriNet, KU Leuven +<br/>imec-DistriNet, KU Leuven +<br/>imec-DistriNet, KU Leuven +<br/>imec-DistriNet, KU Leuven +<br/>imec-DistriNet, KU Leuven +</td></tr><tr><td>b908edadad58c604a1e4b431f69ac8ded350589a</td><td>Deep Face Feature for Face Alignment +</td></tr><tr><td>b9f2a755940353549e55690437eb7e13ea226bbf</td><td>Unsupervised Feature Learning from Videos for Discovering and Recognizing Actions +</td></tr><tr><td>b9cedd1960d5c025be55ade0a0aa81b75a6efa61</td><td>INEXACT KRYLOV SUBSPACE ALGORITHMS FOR LARGE +<br/>MATRIX EXPONENTIAL EIGENPROBLEM FROM +<br/>DIMENSIONALITY REDUCTION +</td></tr><tr><td>b971266b29fcecf1d5efe1c4dcdc2355cb188ab0</td><td>MAI et al.: ON THE RECONSTRUCTION OF FACE IMAGES FROM DEEP FACE TEMPLATES +<br/>On the Reconstruction of Face Images from +<br/>Deep Face Templates +</td></tr><tr><td>a158c1e2993ac90a90326881dd5cb0996c20d4f3</td><td>OPEN ACCESS +<br/>ISSN 2073-8994 +<br/>Article +<br/>1 DMA, Università degli Studi di Palermo, via Archirafi 34, 90123 Palermo, Italy +<br/>2 CITC, Università degli Studi di Palermo, via Archirafi 34, 90123 Palermo, Itlay +<br/>3 Istituto Nazionale di Ricerche Demopolis, via Col. Romey 7, 91100 Trapani, Italy +<br/>† Deceased on 15 March 2009. +<br/>Received: 4 March 2010; in revised form: 23 March 2010 / Accepted: 29 March 2010 / +<br/>Published: 1 April 2010 +</td></tr><tr><td>a15d9d2ed035f21e13b688a78412cb7b5a04c469</td><td>Object Detection Using +<br/>Strongly-Supervised Deformable Part Models +<br/>1Computer Vision and Active Perception Laboratory (CVAP), KTH, Sweden +<br/>2INRIA, WILLOW, Laboratoire d’Informatique de l’Ecole Normale Superieure +</td></tr><tr><td>a1b1442198f29072e907ed8cb02a064493737158</td><td>456 +<br/>Crowdsourcing Facial Responses +<br/>to Online Videos +</td></tr><tr><td>a15c728d008801f5ffc7898568097bbeac8270a4</td><td>Concise Preservation by Combining Managed Forgetting +<br/>and Contextualized Remembering +<br/>Grant Agreement No. 600826 +<br/>Deliverable D4.4 +<br/>Work-package +<br/>Deliverable +<br/>Deliverable Leader +<br/>Quality Assessor +<br/>Dissemination level +<br/>Delivery date in Annex I +<br/>Actual delivery date +<br/>Revisions +<br/>Status +<br/>Keywords +<br/>Information Consolidation and Con- +<br/>WP4: +<br/>centration +<br/>D4.4: +<br/>Information analysis, consolidation +<br/>and concentration techniques, and evalua- +<br/>tion - Final release. +<br/>Vasileios Mezaris (CERTH) +<br/>Walter Allasia (EURIX) +<br/>PU +<br/>31-01-2016 (M36) +<br/>31-01-2016 +<br/>Final +<br/>multidocument summarization, semantic en- +<br/>richment, +<br/>feature extraction, concept de- +<br/>tection, event detection, image/video qual- +<br/>ity, image/video aesthetic quality, face de- +<br/>tection/clustering, +<br/>im- +<br/>age/video summarization, image/video near +<br/>duplicate detection, data deduplication, con- +<br/>densation, consolidation +<br/>image clustering, +</td></tr><tr><td>a1132e2638a8abd08bdf7fc4884804dd6654fa63</td><td>6 +<br/>Real-Time Video Face Recognition +<br/>for Embedded Devices +<br/>Tessera, Galway, +<br/>Ireland +<br/>1. Introduction +<br/>This chapter will address the challenges of real-time video face recognition systems +<br/>implemented in embedded devices. Topics to be covered include: the importance and +<br/>challenges of video face recognition in real life scenarios, describing a general architecture of +<br/>a generic video face recognition system and a working solution suitable for recognizing +<br/>faces in real-time using low complexity devices. Each component of the system will be +<br/>described together with the system’s performance on a database of video samples that +<br/>resembles real life conditions. +<br/>2. Video face recognition +<br/>Face recognition remains a very active topic in computer vision and receives attention from +<br/>a large community of researchers in that discipline. Many reasons feed this interest; the +<br/>main being the wide range of commercial, law enforcement and security applications that +<br/>require authentication. The progress made in recent years on the methods and algorithms +<br/>for data processing as well as the availability of new technologies makes it easier to study +<br/>these algorithms and turn them into commercially viable product. Biometric based security +<br/>systems are becoming more popular due to their non-invasive nature and their increasing +<br/>reliability. Surveillance applications based on face recognition are gaining increasing +<br/>attention after the United States’ 9/11 events and with the ongoing security threats. The +<br/>Face Recognition Vendor Test (FRVT) (Phillips et al., 2003) includes video face recognition +<br/>testing starting with the 2002 series of tests. +<br/>Recently, face recognition technology was deployed in consumer applications such as +<br/>organizing a collection of images using the faces present in the images (Picassa; Corcoran & +<br/>Costache, 2005), prioritizing family members for best capturing conditions when taking +<br/>pictures, or directly annotating the images as they are captured (Costache et al., 2006). +<br/>Video face recognition, compared with more traditional still face recognition, has the main +<br/>advantage of using multiple instances of the same individual in sequential frames for +<br/>recognition to occur. In still recognition case, the system has only one input image to make +<br/>the decision if the person is or is not in the database. If the image is not suitable for +<br/>recognition (due to face orientation, expression, quality or facial occlusions) the recognition +<br/>result will most likely be incorrect. In the video image there are multiple frames which can +<br/>www.intechopen.com +</td></tr><tr><td>a14ae81609d09fed217aa12a4df9466553db4859</td><td>REVISED VERSION, JUNE 2011 +<br/>Face Identification Using Large Feature Sets +</td></tr><tr><td>a1e97c4043d5cc9896dc60ae7ca135782d89e5fc</td><td>IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE +<br/>Re-identification of Humans in Crowds using +<br/>Personal, Social and Environmental Constraints +</td></tr><tr><td>efd308393b573e5410455960fe551160e1525f49</td><td>Tracking Persons-of-Interest via +<br/>Unsupervised Representation Adaptation +</td></tr><tr><td>ef4ecb76413a05c96eac4c743d2c2a3886f2ae07</td><td>Modeling the Importance of Faces in Natural Images +<br/>Jin B.a, Yildirim G.a, Lau C.a, Shaji A.a, Ortiz Segovia M.b and S¨usstrunk S.a +<br/>aEPFL, Lausanne, Switzerland; +<br/>bOc´e, Paris, France +</td></tr><tr><td>ef032afa4bdb18b328ffcc60e2dc5229cc1939bc</td><td>Fang and Yuan EURASIP Journal on Image and Video +<br/>Processing (2018) 2018:44 +<br/>https://doi.org/10.1186/s13640-018-0282-x +<br/>EURASIP Journal on Image +<br/>and Video Processing +<br/>RESEARCH +<br/>Open Access +<br/>Attribute-enhanced metric learning for +<br/>face retrieval +</td></tr><tr><td>ef5531711a69ed687637c48930261769465457f0</td><td>Studio2Shop: from studio photo shoots to fashion articles +<br/>Zalando Research, Muehlenstr. 25, 10243 Berlin, Germany +<br/>Keywords: +<br/>computer vision, deep learning, fashion, item recognition, street-to-shop +</td></tr><tr><td>efa08283656714911acff2d5022f26904e451113</td><td>Active Object Localization in Visual Situations +</td></tr><tr><td>ef999ab2f7b37f46445a3457bf6c0f5fd7b5689d</td><td>Calhoun: The NPS Institutional Archive +<br/>DSpace Repository +<br/>Theses and Dissertations +<br/>1. Thesis and Dissertation Collection, all items +<br/>2017-12 +<br/>Improving face verification in photo albums by +<br/>combining facial recognition and metadata +<br/>with cross-matching +<br/>Monterey, California: Naval Postgraduate School +<br/>http://hdl.handle.net/10945/56868 +<br/>Downloaded from NPS Archive: Calhoun +</td></tr><tr><td>c3beae515f38daf4bd8053a7d72f6d2ed3b05d88</td><td></td></tr><tr><td>c3dc4f414f5233df96a9661609557e341b71670d</td><td>Tao et al. EURASIP Journal on Advances in Signal Processing 2011, 2011:4 +<br/>http://asp.eurasipjournals.com/content/2011/1/4 +<br/>RESEARCH +<br/>Utterance independent bimodal emotion +<br/>recognition in spontaneous communication +<br/>Open Access +</td></tr><tr><td>c398684270543e97e3194674d9cce20acaef3db3</td><td>Chapter 2 +<br/>Comparative Face Soft Biometrics for +<br/>Human Identification +</td></tr><tr><td>c3285a1d6ec6972156fea9e6dc9a8d88cd001617</td><td></td></tr><tr><td>c3418f866a86dfd947c2b548cbdeac8ca5783c15</td><td></td></tr><tr><td>c32383330df27625592134edd72d69bb6b5cff5c</td><td>422 +<br/>IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART B: CYBERNETICS, VOL. 42, NO. 2, APRIL 2012 +<br/>Intrinsic Illumination Subspace for Lighting +<br/>Insensitive Face Recognition +</td></tr><tr><td>c3a3f7758bccbead7c9713cb8517889ea6d04687</td><td></td></tr><tr><td>c30e4e4994b76605dcb2071954eaaea471307d80</td><td></td></tr><tr><td>c37a971f7a57f7345fdc479fa329d9b425ee02be</td><td>A Novice Guide towards Human Motion Analysis and Understanding +</td></tr><tr><td>c3638b026c7f80a2199b5ae89c8fcbedfc0bd8af</td><td></td></tr><tr><td>c3fb2399eb4bcec22723715556e31c44d086e054</td><td>499 +<br/>2014 IEEE International Conference on Acoustic, Speech and Signal Processing (ICASSP) +<br/>978-1-4799-2893-4/14/$31.00 ©2014 IEEE +<br/>1. INTRODUCTION +</td></tr><tr><td>c37de914c6e9b743d90e2566723d0062bedc9e6a</td><td>©2016 Society for Imaging Science and Technology +<br/>DOI: 10.2352/ISSN.2470-1173.2016.11.IMAWM-455 +<br/>Joint and Discriminative Dictionary Learning +<br/>Expression Recognition +<br/>for Facial +</td></tr><tr><td>c4f1fcd0a5cdaad8b920ee8188a8557b6086c1a4</td><td>Int J Comput Vis (2014) 108:3–29 +<br/>DOI 10.1007/s11263-014-0698-4 +<br/>The Ignorant Led by the Blind: A Hybrid Human–Machine Vision +<br/>System for Fine-Grained Categorization +<br/>Received: 7 March 2013 / Accepted: 8 January 2014 / Published online: 20 February 2014 +<br/>© Springer Science+Business Media New York 2014 +</td></tr><tr><td>c4dcf41506c23aa45c33a0a5e51b5b9f8990e8ad</td><td> Understanding Activity: Learning the Language of Action +<br/> Univ. of Rochester and Maryland +<br/>1.1 Overview +<br/>Understanding observed activity is an important +<br/>problem, both from the standpoint of practical applications, +<br/>and as a central issue in attempting to describe the +<br/>phenomenon of intelligence. On the practical side, there are a +<br/>large number of applications that would benefit from +<br/>improved machine ability to analyze activity. The most +<br/>prominent are various surveillance scenarios. The current +<br/>emphasis on homeland security has brought this issue to the +<br/>forefront, and resulted in considerable work on mostly low- +<br/>level detection schemes. There are also applications in +<br/>medical diagnosis and household assistants that, in the long +<br/>run, may be even more important. In addition, there are +<br/>numerous scientific projects, ranging from monitoring of +<br/>weather conditions to observation of animal behavior that +<br/>would be facilitated by automatic understanding of activity. +<br/>From a scientific standpoint, understanding activity +<br/>understanding is central to understanding intelligence. +<br/>Analyzing what is happening in the environment, and acting +<br/>on the results of that analysis is, to a large extent, what +<br/>natural intelligent systems do, whether they are human or +<br/>animal. Artificial intelligences, if we want them to work with +<br/>people in the natural world, will need commensurate abilities. +<br/>The importance of the problem has not gone unrecognized. +<br/>There is a substantial body of work on various components of +<br/>the problem, most especially on change detection, motion +<br/>analysis, and tracking. More recently, in the context of +<br/>surveillance applications, there have been some preliminary +<br/>efforts to come up with a general ontology of human activity. +<br/>These efforts have largely been top-down in the classic AI +<br/>tradition, and, as with earlier analogous effort in areas such +<br/>as object recognition and scene understanding, have seen +<br/>limited practical application because of the difficulty in +<br/>robustly extracting the putative primitives on which the top- +<br/>down formalism is based. We propose a novel alternative +<br/>approach, where understanding activity is centered on +</td></tr><tr><td>c49aed65fcf9ded15c44f9cbb4b161f851c6fa88</td><td>Multiscale Facial Expression Recognition using Convolutional Neural Networks +<br/>IDIAP, Martigny, Switzerland +</td></tr><tr><td>eac6aee477446a67d491ef7c95abb21867cf71fc</td><td>JOURNAL +<br/>A survey of sparse representation: algorithms and +<br/>applications +</td></tr><tr><td>ea482bf1e2b5b44c520fc77eab288caf8b3f367a</td><td>Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence (IJCAI-17) +<br/>2592 +</td></tr><tr><td>eafda8a94e410f1ad53b3e193ec124e80d57d095</td><td>Jeffrey F. Cohn +<br/>13 +<br/>Observer-Based Measurement of Facial Expression +<br/>With the Facial Action Coding System +<br/>Facial expression has been a focus of emotion research for over +<br/>a hundred years (Darwin, 1872/1998). It is central to several +<br/>leading theories of emotion (Ekman, 1992; Izard, 1977; +<br/>Tomkins, 1962) and has been the focus of at times heated +<br/>debate about issues in emotion science (Ekman, 1973, 1993; +<br/>Fridlund, 1992; Russell, 1994). Facial expression figures +<br/>prominently in research on almost every aspect of emotion, +<br/>including psychophysiology (Levenson, Ekman, & Friesen, +<br/>1990), neural bases (Calder et al., 1996; Davidson, Ekman, +<br/>Saron, Senulis, & Friesen, 1990), development (Malatesta, +<br/>Culver, Tesman, & Shephard, 1989; Matias & Cohn, 1993), +<br/>perception (Ambadar, Schooler, & Cohn, 2005), social pro- +<br/>cesses (Hatfield, Cacioppo, & Rapson, 1992; Hess & Kirouac, +<br/>2000), and emotion disorder (Kaiser, 2002; Sloan, Straussa, +<br/>Quirka, & Sajatovic, 1997), to name a few. +<br/>Because of its importance to the study of emotion, a num- +<br/>ber of observer-based systems of facial expression measure- +<br/>ment have been developed (Ekman & Friesen, 1978, 1982; +<br/>Ekman, Friesen, & Tomkins, 1971; Izard, 1979, 1983; Izard +<br/>& Dougherty, 1981; Kring & Sloan, 1991; Tronick, Als, & +<br/>Brazelton, 1980). Of these various systems for describing +<br/>facial expression, the Facial Action Coding System (FACS; +<br/>Ekman & Friesen, 1978; Ekman, Friesen, & Hager, 2002) is +<br/>the most comprehensive, psychometrically rigorous, and +<br/>widely used (Cohn & Ekman, 2005; Ekman & Rosenberg, +<br/>2005). Using FACS and viewing video-recorded facial behav- +<br/>ior at frame rate and slow motion, coders can manually code +<br/>nearly all possible facial expressions, which are decomposed +<br/>into action units (AUs). Action units, with some qualifica- +<br/>tions, are the smallest visually discriminable facial move- +<br/>ments. By comparison, other systems are less thorough +<br/>(Malatesta et al., 1989), fail to differentiate between some +<br/>anatomically distinct movements (Oster, Hegley, & Nagel, +<br/>1992), consider movements that are not anatomically dis- +<br/>tinct as separable (Oster et al., 1992), and often assume a one- +<br/>to-one mapping between facial expression and emotion (for +<br/>a review of these systems, see Cohn & Ekman, in press). +<br/>Unlike systems that use emotion labels to describe ex- +<br/>pression, FACS explicitly distinguishes between facial actions +<br/>and inferences about what they mean. FACS itself is descrip- +<br/>tive and includes no emotion-specified descriptors. Hypoth- +<br/>eses and inferences about the emotional meaning of facial +<br/>actions are extrinsic to FACS. If one wishes to make emo- +<br/>tion-based inferences from FACS codes, a variety of related +<br/>resources exist. These include the FACS Investigators’ Guide +<br/>(Ekman et al., 2002), the FACS interpretive database (Ekman, +<br/>Rosenberg, & Hager, 1998), and a large body of empirical +<br/>research.(Ekman & Rosenberg, 2005). These resources sug- +<br/>gest combination rules for defining emotion-specified expres- +<br/>sions from FACS action units, but this inferential step remains +<br/>extrinsic to FACS. Because of its descriptive power, FACS +<br/>is regarded by many as the standard measure for facial be- +<br/>havior and is used widely in diverse fields. Beyond emo- +<br/>tion science, these include facial neuromuscular disorders +<br/>(Van Swearingen & Cohn, 2005), neuroscience (Bruce & +<br/>Young, 1998; Rinn, 1984, 1991), computer vision (Bartlett, +<br/>203 +<br/>UNPROOFED PAGES </td></tr><tr><td>ea85378a6549bb9eb9bcc13e31aa6a61b655a9af</td><td>Diplomarbeit +<br/>Template Protection for PCA-LDA-based 3D +<br/>Face Recognition System +<br/>von +<br/>Technische Universität Darmstadt +<br/>Fachbereich Informatik +<br/>Fachgebiet Graphisch-Interaktive Systeme +<br/>Fraunhoferstraße 5 +<br/>64283 Darmstadt +</td></tr><tr><td>ea2ee5c53747878f30f6d9c576fd09d388ab0e2b</td><td>Viola-Jones based Detectors: How much affects +<br/>the Training Set? +<br/>SIANI +<br/>Edif. Central del Parque Cient´ıfico Tecnol´ogico +<br/>Universidad de Las Palmas de Gran Canaria +<br/>35017 - Spain +</td></tr><tr><td>ea96bc017fb56593a59149e10d5f14011a3744a0</td><td></td></tr><tr><td>e10a257f1daf279e55f17f273a1b557141953ce2</td><td></td></tr><tr><td>e171fba00d88710e78e181c3e807c2fdffc6798a</td><td></td></tr><tr><td>e1ab3b9dee2da20078464f4ad8deb523b5b1792e</td><td>Pre-Training CNNs Using Convolutional +<br/>Autoencoders +<br/>TU Berlin +<br/>TU Berlin +<br/>Sabbir Ahmmed +<br/>TU Berlin +<br/>TU Berlin +</td></tr><tr><td>e16efd2ae73a325b7571a456618bfa682b51aef8</td><td></td></tr><tr><td>e19ebad4739d59f999d192bac7d596b20b887f78</td><td>Learning Gating ConvNet for Two-Stream based Methods in Action +<br/>Recognition +</td></tr><tr><td>e13360cda1ebd6fa5c3f3386c0862f292e4dbee4</td><td></td></tr><tr><td>e1d726d812554f2b2b92cac3a4d2bec678969368</td><td>J Electr Eng Technol.2015; 10(?): 30-40 +<br/>http://dx.doi.org/10.5370/JEET.2015.10.2.030 +<br/>ISSN(Print) +<br/>1975-0102 +<br/>ISSN(Online) 2093-7423 +<br/>Human Action Recognition Bases on Local Action Attributes +<br/>and Mohan S Kankanhalli** +</td></tr><tr><td>e1e6e6792e92f7110e26e27e80e0c30ec36ac9c2</td><td>TSINGHUA SCIENCE AND TECHNOLOGY +<br/>ISSNll1007-0214 +<br/>0?/?? pp???–??? +<br/>DOI: 10.26599/TST.2018.9010000 +<br/>Volume 1, Number 1, Septembelr 2018 +<br/>Ranking with Adaptive Neighbors +</td></tr><tr><td>cd9666858f6c211e13aa80589d75373fd06f6246</td><td>A Novel Time Series Kernel for +<br/>Sequences Generated by LTI Systems +<br/>V.le delle Scienze Ed.6, DIID, Universit´a degli studi di Palermo, Italy +</td></tr><tr><td>cd4c047f4d4df7937aff8fc76f4bae7718004f40</td><td></td></tr><tr><td>cd596a2682d74bdfa7b7160dd070b598975e89d9</td><td>Mood Detection: Implementing a facial +<br/>expression recognition system +<br/>1. Introduction +<br/>Facial expressions play a significant role in human dialogue. As a result, there has been +<br/>considerable work done on the recognition of emotional expressions and the application of this +<br/>research will be beneficial in improving human-machine dialogue. One can imagine the +<br/>improvements to computer interfaces, automated clinical (psychological) research or even +<br/>interactions between humans and autonomous robots. +<br/>Unfortunately, a lot of the literature does not focus on trying to achieve high recognition rates +<br/>across multiple databases. In this project we develop our own mood detection system that +<br/>addresses this challenge. The system involves pre-processing image data by normalizing and +<br/>applying a simple mask, extracting certain (facial) features using PCA and Gabor filters and then +<br/>using SVMs for classification and recognition of expressions. Eigenfaces for each class are used +<br/>to determine class-specific masks which are then applied to the image data and used to train +<br/>multiple, one against the rest, SVMs. We find that simply using normalized pixel intensities +<br/>works well with such an approach. +<br/>Figure 1 – Overview of our system design +<br/>2. Image pre-processing +<br/>We performed pre-processing on the images used to train and test our algorithms as follows: +<br/>1. The location of the eyes is first selected manually +<br/>2. Images are scaled and cropped to a fixed size (170 x 130) keeping the eyes in all images +<br/>aligned +<br/>3. The image is histogram equalized using the mean histogram of all the training images to +<br/>make it invariant to lighting, skin color etc. +<br/>4. A fixed oval mask is applied to the image to extract face region. This serves to eliminate +<br/>the background, hair, ears and other extraneous features in the image which provide no +<br/>information about facial expression. +<br/>This approach works reasonably well in capturing expression-relevant facial information across +<br/>all databases. Examples of pre-processed images from the various datasets are shown in Figure- +<br/>2a below. +</td></tr><tr><td>cda4fb9df653b5721ad4fe8b4a88468a410e55ec</td><td>Gabor wavelet transform and its application +</td></tr><tr><td>cd3005753012409361aba17f3f766e33e3a7320d</td><td>Multilinear Biased Discriminant Analysis: A Novel Method for Facial +<br/>Action Unit Representation +</td></tr><tr><td>cd7a7be3804fd217e9f10682e0c0bfd9583a08db</td><td>Women also Snowboard: +<br/>Overcoming Bias in Captioning Models +</td></tr><tr><td>ccfcbf0eda6df876f0170bdb4d7b4ab4e7676f18</td><td>JOURNAL OF LATEX CLASS FILES, VOL. 6, NO. 1, JUNE 2011 +<br/>A Dynamic Appearance Descriptor Approach to +<br/>Facial Actions Temporal Modelling +</td></tr><tr><td>ccbfc004e29b3aceea091056b0ec536e8ea7c47e</td><td></td></tr><tr><td>cc3c273bb213240515147e8be68c50f7ea22777c</td><td>Gaining Insight Into Films +<br/>Via Topic Modeling & Visualization +<br/>KEYWORDS Collaboration, computer vision, cultural +<br/>analytics, economy of abundance, interactive data +<br/>visualization +<br/>We moved beyond misuse when the software actually +<br/>became useful for film analysis with the addition of audio +<br/>analysis, subtitle analysis, facial recognition, and topic +<br/>modeling. Using multiple types of visualizations and +<br/>a back-and-fourth workflow between people and AI +<br/>we arrived at an approach for cultural analytics that +<br/>can be used to review and develop film criticism. Finally, +<br/>we present ways to apply these techniques to Database +<br/>Cinema and other aspects of film and video creation. +<br/>PROJECT DATE 2014 +<br/>URL http://misharabinovich.com/soyummy.html +</td></tr><tr><td>cc8e378fd05152a81c2810f682a78c5057c8a735</td><td>International Journal of Computer Sciences and Engineering Open Access +<br/> Research Paper Volume-5, Issue-12 E-ISSN: 2347-2693 +<br/>Expression Invariant Face Recognition System based on Topographic +<br/>Independent Component Analysis and Inner Product Classifier +<br/> +<br/>Department of Electrical Engineering, IIT Delhi, New Delhi, India +<br/>Available online at: www.ijcseonline.org +<br/>Received: 07/Nov/2017, Revised: 22/Nov/2017, Accepted: 14/Dec/2017, Published: 31/Dec/2017 +</td></tr><tr><td>cc31db984282bb70946f6881bab741aa841d3a7c</td><td>ALBANIE, VEDALDI: LEARNING GRIMACES BY WATCHING TV +<br/>Learning Grimaces by Watching TV +<br/>http://www.robots.ox.ac.uk/~albanie +<br/>http://www.robots.ox.ac.uk/~vedaldi +<br/>Engineering Science Department +<br/>Univeristy of Oxford +<br/>Oxford, UK +</td></tr><tr><td>cc8bf03b3f5800ac23e1a833447c421440d92197</td><td></td></tr><tr><td>cc96eab1e55e771e417b758119ce5d7ef1722b43</td><td>An Empirical Study of Recent +<br/>Face Alignment Methods +</td></tr><tr><td>e64b683e32525643a9ddb6b6af8b0472ef5b6a37</td><td>Face Recognition and Retrieval in Video +</td></tr><tr><td>e6b45d5a86092bbfdcd6c3c54cda3d6c3ac6b227</td><td>Pairwise Relational Networks for Face +<br/>Recognition +<br/>1 Department of Creative IT Engineering, POSTECH, Korea +<br/>2 Department of Computer Science and Engineering, POSTECH, Korea +</td></tr><tr><td>e6865b000cf4d4e84c3fe895b7ddfc65a9c4aaec</td><td>Chapter 15. The critical role of the +<br/>cold-start problem and incentive systems +<br/>in emotional Web 2.0 services +</td></tr><tr><td>e6dc1200a31defda100b2e5ddb27fb7ecbbd4acd</td><td>1921 +<br/>Flexible Manifold Embedding: A Framework +<br/>for Semi-Supervised and Unsupervised +<br/>Dimension Reduction +<br/>0 = +<br/>, the linear regression function ( +</td></tr><tr><td>e6e5a6090016810fb902b51d5baa2469ae28b8a1</td><td>Title +<br/>Energy-Efficient Deep In-memory Architecture for NAND +<br/>Flash Memories +<br/>Archived version +<br/>Accepted manuscript: the content is same as the published +<br/>paper but without the final typesetting by the publisher +<br/>Published version +<br/>DOI +<br/>Published paper +<br/>URL +<br/>Authors (contact) +<br/>10.1109/ISCAS.2018.8351458 +</td></tr><tr><td>e6540d70e5ffeed9f447602ea3455c7f0b38113e</td><td></td></tr><tr><td>e6ee36444038de5885473693fb206f49c1369138</td><td></td></tr><tr><td>f913bb65b62b0a6391ffa8f59b1d5527b7eba948</td><td></td></tr><tr><td>f96bdd1e2a940030fb0a89abbe6c69b8d7f6f0c1</td><td></td></tr><tr><td>f0cee87e9ecedeb927664b8da44b8649050e1c86</td><td></td></tr><tr><td>f0f4f16d5b5f9efe304369120651fa688a03d495</td><td>Temporal Generative Adversarial Nets +<br/>Preferred Networks inc., Japan +</td></tr><tr><td>f06b015bb19bd3c39ac5b1e4320566f8d83a0c84</td><td></td></tr><tr><td>f0a3f12469fa55ad0d40c21212d18c02be0d1264</td><td>Sparsity Sharing Embedding for Face +<br/>Verification +<br/>Department of Electrical Engineering, KAIST, Daejeon, Korea +</td></tr><tr><td>f7dea4454c2de0b96ab5cf95008ce7144292e52a</td><td></td></tr><tr><td>f7b422df567ce9813926461251517761e3e6cda0</td><td>FACE AGING WITH CONDITIONAL GENERATIVE ADVERSARIAL NETWORKS +<br/>(cid:63) Orange Labs, 4 rue Clos Courtel, 35512 Cesson-S´evign´e, France +<br/>† Eurecom, 450 route des Chappes, 06410 Biot, France +</td></tr><tr><td>f79c97e7c3f9a98cf6f4a5d2431f149ffacae48f</td><td>Provided by the author(s) and NUI Galway in accordance with publisher policies. Please cite the published +<br/>version when available. +<br/>Title +<br/>On color texture normalization for active appearance models +<br/>Author(s) +<br/>Ionita, Mircea C.; Corcoran, Peter M.; Buzuloiu, Vasile +<br/>Publication +<br/>Date +<br/>2009-05-12 +<br/>Publication +<br/>Information +<br/>Ionita, M. C., Corcoran, P., & Buzuloiu, V. (2009). On Color +<br/>Texture Normalization for Active Appearance Models. Image +<br/>Processing, IEEE Transactions on, 18(6), 1372-1378. +<br/>Publisher +<br/>IEEE +<br/>Link to +<br/>publisher's +<br/>version +<br/>http://dx.doi.org/10.1109/TIP.2009.2017163 +<br/>Item record +<br/>http://hdl.handle.net/10379/1350 +<br/>Some rights reserved. For more information, please see the item record link above. +<br/>Downloaded 2017-06-17T22:38:27Z +</td></tr><tr><td>f7452a12f9bd927398e036ea6ede02da79097e6e</td><td></td></tr><tr><td>f7dcadc5288653ec6764600c7c1e2b49c305dfaa</td><td>Copyright +<br/>by +<br/>Adriana Ivanova Kovashka +<br/>2014 +</td></tr><tr><td>f7de943aa75406fe5568fdbb08133ce0f9a765d4</td><td>Project 1.5: Human Identification at a Distance - Hornak, Adjeroh, Cukic, Gautum, & Ross +<br/>Project 1.5 +<br/>Biometric Identification and Surveillance1 +<br/>Year 5 Deliverable +<br/>Technical Report: +<br/>and +<br/>Research Challenges in Biometrics +<br/>Indexed biography of relevant biometric research literature +<br/>Donald Adjeroh, Bojan Cukic, Arun Ross +<br/>April, 2014 +<br/> +<br/>1 "This research was supported by the United States Department of Homeland Security through the National Center for Border Security +<br/>and Immigration (BORDERS) under grant number 2008-ST-061-BS0002. However, any opinions, findings, and conclusions or +<br/>recommendations in this document are those of the authors and do not necessarily reflect views of the United States Department of +<br/>Homeland Security." +</td></tr><tr><td>f75852386e563ca580a48b18420e446be45fcf8d</td><td>ILLUMINATION INVARIANT FACE RECOGNITION +<br/> +<br/> +<br/> +<br/> +<br/> +<br/> +<br/> +<br/> +<br/> +<br/> +<br/> +<br/> +<br/> +<br/> +<br/> +<br/> +<br/> +<br/>ENEE 631: Digital Image and Video Processing +<br/>Instructor: Dr. K. J. Ray Liu +<br/>Term Project - Spring 2006 +<br/>1. +<br/>INTRODUCTION +<br/> +<br/> +<br/>The performance of the Face Recognition algorithms is severely affected by two +<br/>important factors: the change in Pose and Illumination conditions of the subjects. The +<br/>changes in Illumination conditions of the subjects can be so drastic that, the variation in +<br/>lighting will be of the similar order as that of the variation due to the change in subjects +<br/>[1] and this can result in misclassification. +<br/> +<br/> For example, in the acquisition of the face of a person from a real time video, the +<br/>ambient conditions will cause different lighting variations on the tracked face. Some +<br/>examples of images with different illumination conditions are shown in Fig. 1. In this +<br/>project, we study some algorithms that are capable of performing Illumination Invariant +<br/>Face Recognition. The performances of these algorithms were compared on the CMU- +<br/>Illumination dataset [13], by using the entire face as the input to the algorithms. Then, a +<br/>model of dividing the face into four regions is proposed and the performance of the +<br/>algorithms on these new features is analyzed. +<br/> +<br/> +</td></tr><tr><td>f78863f4e7c4c57744715abe524ae4256be884a9</td><td></td></tr><tr><td>f77c9bf5beec7c975584e8087aae8d679664a1eb</td><td>Local Deep Neural Networks for Age and Gender Classification +<br/>March 27, 2017 +</td></tr><tr><td>e8410c4cd1689829c15bd1f34995eb3bd4321069</td><td></td></tr><tr><td>e8fdacbd708feb60fd6e7843b048bf3c4387c6db</td><td>Deep Learning +<br/>Hinnerup Net A/S +<br/>www.hinnerup.net +<br/>July 4, 2014 +<br/>Introduction +<br/>Deep learning is a topic in the field of artificial intelligence (AI) and is a relatively +<br/>new research area although based on the popular artificial neural networks (supposedly +<br/>mirroring brain function). With the development of the perceptron in the 1950s and +<br/>1960s by Frank RosenBlatt, research began on artificial neural networks. To further +<br/>mimic the architectural depth of the brain, researchers wanted to train a deep multi- +<br/>layer neural network – this, however, did not happen until Geoffrey Hinton in 2006 +<br/>introduced Deep Belief Networks [1]. +<br/>Recently, the topic of deep learning has gained public interest. Large web companies such +<br/>as Google and Facebook have a focused research on AI and an ever increasing amount +<br/>of compute power, which has led to researchers finally being able to produce results +<br/>that are of interest to the general public. In July 2012 Google trained a deep learning +<br/>network on YouTube videos with the remarkable result that the network learned to +<br/>recognize humans as well as cats [6], and in January this year Google successfully used +<br/>deep learning on Street View images to automatically recognize house numbers with +<br/>an accuracy comparable to that of a human operator [5]. In March this year Facebook +<br/>announced their DeepFace algorithm that is able to match faces in photos with Facebook +<br/>users almost as accurately as a human can do [9]. +<br/>Deep learning and other AI are here to stay and will become more and more present in +<br/>our daily lives, so we had better make ourselves acquainted with the technology. Let’s +<br/>dive into the deep water and try not to drown! +<br/>Data Representations +<br/>Before presenting data to an AI algorithm, we would normally prepare the data to make +<br/>it feasible to work with. For instance, if the data consists of images, we would take each +</td></tr><tr><td>e8b2a98f87b7b2593b4a046464c1ec63bfd13b51</td><td>CMS-RCNN: Contextual Multi-Scale +<br/>Region-based CNN for Unconstrained Face +<br/>Detection +</td></tr><tr><td>e8c6c3fc9b52dffb15fe115702c6f159d955d308</td><td>13 +<br/>Linear Subspace Learning for +<br/>Facial Expression Analysis +<br/>Philips Research +<br/>The Netherlands +<br/>1. Introduction +<br/>Facial expression, resulting from movements of the facial muscles, is one of the most +<br/>powerful, natural, and immediate means for human beings to communicate their emotions +<br/>and intentions. Some examples of facial expressions are shown in Fig. 1. Darwin (1872) was +<br/>the first to describe in detail the specific facial expressions associated with emotions in +<br/>animals and humans; he argued that all mammals show emotions reliably in their faces. +<br/>Psychological studies (Mehrabian, 1968; Ambady & Rosenthal, 1992) indicate that facial +<br/>expressions, with other non-verbal cues, play a major and fundamental role in face-to-face +<br/>communication. +<br/>Fig. 1. Facial expressions of George W. Bush. +<br/>Machine analysis of facial expressions, enabling computers to analyze and interpret facial +<br/>expressions as humans do, has many important applications including intelligent human- +<br/>computer interaction, computer animation, surveillance and security, medical diagnosis, +<br/>law enforcement, and awareness system (Shan, 2007). Driven by its potential applications +<br/>and theoretical interests of cognitive and psychological scientists, automatic facial +<br/>expression analysis has attracted much attention in last two decades (Pantic & Rothkrantz, +<br/>2000a; Fasel & Luettin, 2003; Tian et al, 2005; Pantic & Bartlett, 2007). It has been studied in +<br/>multiple disciplines such as psychology, cognitive science, computer vision, pattern +<br/>Source: Machine Learning, Book edited by: Abdelhamid Mellouk and Abdennacer Chebira, +<br/> ISBN 978-3-902613-56-1, pp. 450, February 2009, I-Tech, Vienna, Austria +<br/>www.intechopen.com +</td></tr><tr><td>fab83bf8d7cab8fe069796b33d2a6bd70c8cefc6</td><td>Draft: Evaluation Guidelines for Gender +<br/>Classification and Age Estimation +<br/>July 1, 2011 +<br/>Introduction +<br/>In previous research on gender classification and age estimation did not use a +<br/>standardised evaluation procedure. This makes comparison the different ap- +<br/>proaches difficult. +<br/>Thus we propose here a benchmarking and evaluation protocol for gender +<br/>classification as well as age estimation to set a common ground for future re- +<br/>search in these two areas. +<br/>The evaluations are designed such that there is one scenario under controlled +<br/>labratory conditions and one under uncontrolled real life conditions. +<br/>The datasets were selected with the criteria of being publicly available for +<br/>research purposes. +<br/>File lists for the folds corresponding to the individual benchmarking proto- +<br/>cols will be provided over our website at http://face.cs.kit.edu/befit. We +<br/>will provide two kinds of folds for each of the tasks and conditions: one set of +<br/>folds using the whole dataset and one set of folds using a reduced dataset, which +<br/>is approximately balanced in terms of age, gender and ethnicity. +<br/>2 Gender Classification +<br/>In this task the goal is to determine the gender of the persons depicted in the +<br/>individual images. +<br/>2.1 Data +<br/>In previous works one of the most commonly used databases is the Feret database [1, +<br/>2]. We decided here not to take this database, because of its low number of im- +<br/>ages. +</td></tr><tr><td>fa08a4da5f2fa39632d90ce3a2e1688d147ece61</td><td>Supplementary material for +<br/>“Unsupervised Creation of Parameterized Avatars” +<br/>1 Summary of Notations +<br/>Tab. 1 itemizes the symbols used in the submission. Fig. 2,3,4 of the main text illustrate many of these +<br/>symbols. +<br/>2 DANN results +<br/>Fig. 1 shows side by side samples of the original image and the emoji generated by the method of [1]. +<br/>As can be seen, these results do not preserve the identity very well, despite considerable effort invested in +<br/>finding suitable architectures. +<br/>3 Multiple Images Per Person +<br/>Following [4], we evaluate the visual quality that is obtained per person and not just per image, by testing +<br/>TOS on the Facescrub dataset [3]. For each person p, we considered the set of their images Xp, and selected +<br/>the emoji that was most similar to their source image, i.e., the one for which: +<br/>||f (x) − f (e(c(G(x))))||. +<br/>argmin +<br/>x∈Xp +<br/>(1) +<br/>Fig. 2 depicts the results obtained by this selection method on sample images form the Facescrub dataset +<br/>(it is an extension of Fig. 7 of the main text). The figure also shows, for comparison, the DTN [4] result for +<br/>the same image. +<br/>4 Detailed Architecture of the Various Networks +<br/>In this section we describe the architectures of the networks used in for the emoji and avatar experiments. +<br/>4.1 TOS +<br/>Network g maps DeepFace’s 256-dimensional representation [5] into 64 × 64 RGB emoji images. Follow- +<br/>ing [4], this is done through a network with 9 blocks, each consisting of a convolution, batch-normalization +<br/>and ReLU, except the last layer which employs Tanh activation. The odd blocks 1,3,5,7,9 perform upscaling +<br/>convolutions with 512-256-128-64-3 filters respectively of spatial size 4 × 4. The even ones perform 1 × 1 +<br/>convolutions [2]. The odd blocks use a stride of 2 and padding of 1, excluding the first one which does not +<br/>use stride or padding. +<br/>Network e maps emoji parameterization into the matching 64× 64 RGB emoji. The parameterization is +<br/>given as binary vectors in R813 for emojis; Avatar parameterization is in R354. While there are dependencies +<br/>among the various dimensions (an emoji cannot have two hairstyles at once), the binary representation is +<br/>chosen for its simplicity and generality. e is trained in a fully supervised way, using pairs of matching +<br/>parameterization vectors and images in a supervised manner. +<br/>The architecture of e employs five upscaling convolutions with 512-256-128-64-3 filters respectively, +<br/>each of spatial size 4×4. All layers except the last one are batch normalized followed by a ReLU activation. +<br/>The last layer is followed by Tanh activation, generating an RGB image with values in range [−1, 1]. All +<br/>the layers use a stride of 2 and padding of 1, excluding the first one which does not use stride or padding. +</td></tr><tr><td>faead8f2eb54c7bc33bc7d0569adc7a4c2ec4c3b</td><td></td></tr><tr><td>faf5583063682e70dedc4466ac0f74eeb63169e7</td><td></td></tr><tr><td>fad895771260048f58d12158a4d4d6d0623f4158</td><td>Audio-Visual Emotion +<br/>Recognition For Natural +<br/>Human-Robot Interaction +<br/>Dissertation zur Erlangung des akademischen Grades +<br/>Doktor der Ingenieurwissenschaften (Dr.-Ing.) +<br/>vorgelegt von +<br/>an der Technischen Fakultät der Universität Bielefeld +<br/>15. März 2010 +</td></tr><tr><td>ff8315c1a0587563510195356c9153729b533c5b</td><td>432 +<br/>Zapping Index:Using Smile to Measure +<br/>Advertisement Zapping Likelihood +</td></tr><tr><td>ff44d8938c52cfdca48c80f8e1618bbcbf91cb2a</td><td>Towards Video Captioning with Naming: a +<br/>Novel Dataset and a Multi-Modal Approach +<br/>Dipartimento di Ingegneria “Enzo Ferrari” +<br/>Universit`a degli Studi di Modena e Reggio Emilia +</td></tr><tr><td>fffefc1fb840da63e17428fd5de6e79feb726894</td><td>Fine-Grained Age Estimation in the wild with +<br/>Attention LSTM Networks +</td></tr><tr><td>ff398e7b6584d9a692e70c2170b4eecaddd78357</td><td></td></tr><tr><td>ffd81d784549ee51a9b0b7b8aaf20d5581031b74</td><td>Performance Analysis of Retina and DoG +<br/>Filtering Applied to Face Images for Training +<br/>Correlation Filters +<br/>Everardo Santiago Ram(cid:19)(cid:16)rez1, Jos(cid:19)e (cid:19)Angel Gonz(cid:19)alez Fraga1, Omar (cid:19)Alvarez +<br/>1 Facultad de Ciencias, Universidad Aut(cid:19)onoma de Baja California, +<br/>Carretera Transpeninsular Tijuana-Ensenada, N(cid:19)um. 3917, Colonia Playitas, +<br/>Ensenada, Baja California, C.P. 22860 +<br/>{everardo.santiagoramirez,angel_fraga, +<br/>2 Facultad de Ingenier(cid:19)(cid:16)a, Arquitectura y Dise~no, Universidad Aut(cid:19)onoma de Baja +<br/>California, Carretera Transpeninsular Tijuana-Ensenada, N(cid:19)um. 3917, Colonia +<br/>Playitas, Ensenada, Baja California, C.P. 22860 +</td></tr><tr><td>ff60d4601adabe04214c67e12253ea3359f4e082</td><td></td></tr><tr><td>ff8ef43168b9c8dd467208a0b1b02e223b731254</td><td>BreakingNews: Article Annotation by +<br/>Image and Text Processing +</td></tr><tr><td>ffcbedb92e76fbab083bb2c57d846a2a96b5ae30</td><td></td></tr><tr><td>c50d73557be96907f88b59cfbd1ab1b2fd696d41</td><td>JournalofElectronicImaging13(3),474–485(July2004). +<br/>Semiconductor sidewall shape estimation +<br/>Oak Ridge National Laboratory +<br/>Oak Ridge, Tennessee 37831-6010 +</td></tr><tr><td>c54f9f33382f9f656ec0e97d3004df614ec56434</td><td></td></tr><tr><td>c574c72b5ef1759b7fd41cf19a9dcd67e5473739</td><td>Zlatintsi et al. EURASIP Journal on Image and Video Processing (2017) 2017:54 +<br/>DOI 10.1186/s13640-017-0194-1 +<br/>EURASIP Journal on Image +<br/>and Video Processing +<br/>RESEARCH +<br/>Open Access +<br/>COGNIMUSE: a multimodal video +<br/>database annotated with saliency, events, +<br/>semantics and emotion with application to +<br/>summarization +</td></tr><tr><td>c5a561c662fc2b195ff80d2655cc5a13a44ffd2d</td><td>Using Language to Learn Structured Appearance +<br/>Models for Image Annotation +</td></tr><tr><td>c5fe40875358a286594b77fa23285fcfb7bda68e</td><td></td></tr><tr><td>c5be0feacec2860982fbbb4404cf98c654142489</td><td>Semi-Qualitative Probabilistic Networks in Computer +<br/>Vision Problems +<br/>Troy, NY 12180, USA. +<br/>Troy, NY 12180, USA. +<br/>Troy, NY 12180, USA. +<br/>Troy, NY 12180, USA. +<br/>Received: *** +<br/>Revised: *** +</td></tr><tr><td>c5844de3fdf5e0069d08e235514863c8ef900eb7</td><td>Lam S K et al. / (IJCSE) International Journal on Computer Science and Engineering +<br/>Vol. 02, No. 08, 2010, 2659-2665 +<br/>A Study on Similarity Computations in Template +<br/>Matching Technique for Identity Verification +<br/>Lam, S. K., Yeong, C. Y., Yew, C. T., Chai, W. S., Suandi, S. A. +<br/>Intelligent Biometric Group, School of Electrical and Electronic Engineering +<br/>Engineering Campus, Universiti Sains Malaysia +<br/>14300 Nibong Tebal, Pulau Pinang, MALAYSIA +</td></tr><tr><td>c220f457ad0b28886f8b3ef41f012dd0236cd91a</td><td>JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 +<br/>Crystal Loss and Quality Pooling for +<br/>Unconstrained Face Verification and Recognition +</td></tr><tr><td>c254b4c0f6d5a5a45680eb3742907ec93c3a222b</td><td>A Fusion-based Gender Recognition Method +<br/>Using Facial Images +</td></tr><tr><td>c28461e266fe0f03c0f9a9525a266aa3050229f0</td><td>Automatic Detection of Facial Feature Points via +<br/>HOGs and Geometric Prior Models +<br/>1 Computer Vision Center , Universitat Aut`onoma de Barcelona +<br/>2 Universitat Oberta de Catalunya +<br/>3 Dept. de Matem`atica Aplicada i An`alisi +<br/>Universitat de Barcelona +</td></tr><tr><td>c29e33fbd078d9a8ab7adbc74b03d4f830714cd0</td><td></td></tr><tr><td>f68ed499e9d41f9c3d16d843db75dc12833d988d</td><td></td></tr><tr><td>f6ca29516cce3fa346673a2aec550d8e671929a6</td><td>International Journal of Engineering and Advanced Technology (IJEAT) +<br/>ISSN: 2249 – 8958, Volume-2, Issue-4, April 2013 +<br/>Algorithm for Face Matching Using Normalized +<br/>Cross-Correlation +<br/> +</td></tr><tr><td>f6c70635241968a6d5fd5e03cde6907022091d64</td><td></td></tr><tr><td>f6ce34d6e4e445cc2c8a9b8ba624e971dd4144ca</td><td>Cross-label Suppression: A Discriminative and Fast +<br/>Dictionary Learning with Group Regularization +<br/>April 24, 2017 +</td></tr><tr><td>f6abecc1f48f6ec6eede4143af33cc936f14d0d0</td><td></td></tr><tr><td>f6fa97fbfa07691bc9ff28caf93d0998a767a5c1</td><td>k2-means for fast and accurate large scale clustering +<br/>Computer Vision Lab +<br/>D-ITET +<br/>ETH Zurich +<br/>Computer Vision Lab +<br/>D-ITET +<br/>ETH Zurich +<br/>ESAT, KU Leuven +<br/>D-ITET, ETH Zurich +</td></tr><tr><td>e9ed17fd8bf1f3d343198e206a4a7e0561ad7e66</td><td>International Journal of Enhanced Research in Science Technology & Engineering, ISSN: 2319-7463 +<br/>Vol. 3 Issue 1, January-2014, pp: (362-365), Impact Factor: 1.252, Available online at: www.erpublications.com +<br/>Cognitive Learning for Social Robot through +<br/>Facial Expression from Video Input +<br/>1Department of Automation & Robotics, 2Department of Computer Science & Engg. +</td></tr><tr><td>e988be047b28ba3b2f1e4cdba3e8c94026139fcf</td><td>Multi-Task Convolutional Neural Network for +<br/>Pose-Invariant Face Recognition +</td></tr><tr><td>e9d43231a403b4409633594fa6ccc518f035a135</td><td>Deformable Part Models with CNN Features +<br/>Kokkinos1,2 +<br/>1 Ecole Centrale Paris,2 INRIA, 3TTI-Chicago (cid:63) +</td></tr><tr><td>e9fcd15bcb0f65565138dda292e0c71ef25ea8bb</td><td>Repositorio Institucional de la Universidad Autónoma de Madrid +<br/>https://repositorio.uam.es +<br/>Esta es la versión de autor de la comunicación de congreso publicada en: +<br/>This is an author produced version of a paper published in: +<br/>Highlights on Practical Applications of Agents and Multi-Agent Systems: +<br/>International Workshops of PAAMS. Communications in Computer and +<br/>Information Science, Volumen 365. Springer, 2013. 223-230 +<br/>DOI: http://dx.doi.org/10.1007/978-3-642-38061-7_22 +<br/>Copyright: © 2013 Springer-Verlag +<br/>El acceso a la versión del editor puede requerir la suscripción del recurso +<br/>Access to the published version may require subscription +</td></tr><tr><td>e9363f4368b04aeaa6d6617db0a574844fc59338</td><td>BENCHIP: Benchmarking Intelligence +<br/>Processors +<br/>1ICT CAS,2Cambricon,3Alibaba Infrastructure Service, Alibaba Group +<br/>4IFLYTEK,5JD,6RDA Microelectronics,7AMD +</td></tr><tr><td>f16a605abb5857c39a10709bd9f9d14cdaa7918f</td><td>Fast greyscale road sign model matching +<br/>and recognition +<br/>Centre de Visió per Computador +<br/>Edifici O – Campus UAB, 08193 Bellaterra, Barcelona, Catalonia, Spain +</td></tr><tr><td>f1748303cc02424704b3a35595610890229567f9</td><td></td></tr><tr><td>f19ab817dd1ef64ee94e94689b0daae0f686e849</td><td>TECHNISCHE UNIVERSIT¨AT M ¨UNCHEN +<br/>Lehrstuhl f¨ur Mensch-Maschine-Kommunikation +<br/>Blickrichtungsunabh¨angige Erkennung von +<br/>Personen in Bild- und Tiefendaten +<br/>Andre St¨ormer +<br/>Vollst¨andiger Abdruck der von der Fakult¨at f¨ur Elektrotechnik und Informationstechnik +<br/>der Technischen Universit¨at M¨unchen zur Erlangung des akademischen Grades eines +<br/>Doktor-Ingenieurs (Dr.-Ing.) +<br/>genehmigten Dissertation. +<br/>Vorsitzender: +<br/>Univ.-Prof. Dr.-Ing. Thomas Eibert +<br/>Pr¨ufer der Dissertation: +<br/>1. Univ.-Prof. Dr.-Ing. habil. Gerhard Rigoll +<br/>2. Univ.-Prof. Dr.-Ing. Horst-Michael Groß, +<br/>Technische Universit¨at Ilmenau +<br/>Die Dissertation wurde am 16.06.2009 bei der Technischen Universit¨at M¨unchen einge- +<br/>reicht und durch die Fakult¨at f¨ur Elektrotechnik und Informationstechnik am 30.10.2009 +<br/>angenommen. +</td></tr><tr><td>e76798bddd0f12ae03de26b7c7743c008d505215</td><td></td></tr><tr><td>e726acda15d41b992b5a41feabd43617fab6dc23</td><td></td></tr><tr><td>e7b6887cd06d0c1aa4902335f7893d7640aef823</td><td>Modelling of Facial Aging and Kinship: A Survey +</td></tr><tr><td>cb004e9706f12d1de83b88c209ac948b137caae0</td><td>Face Aging Effect Simulation using Hidden Factor +<br/>Analysis Joint Sparse Representation +</td></tr><tr><td>cb9092fe74ea6a5b2bb56e9226f1c88f96094388</td><td></td></tr><tr><td>cb08f679f2cb29c7aa972d66fe9e9996c8dfae00</td><td>JOURNAL OF LATEX CLASS FILES, VOL. 13, NO. 9, SEPTEMBER 2014 +<br/>Action Understanding +<br/>with Multiple Classes of Actors +</td></tr><tr><td>cb84229e005645e8623a866d3d7956c197f85e11</td><td>IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. X, NO. X, MONTH 201X +<br/>Disambiguating Visual Verbs +</td></tr><tr><td>cbe859d151466315a050a6925d54a8d3dbad591f</td><td>GAZE SHIFTS AS DYNAMICAL RANDOM SAMPLING +<br/>Dipartimento di Scienze dell’Informazione +<br/>Universit´a di Milano +<br/>Via Comelico 39/41 +<br/>20135 Milano, Italy +</td></tr><tr><td>f8c94afd478821681a1565d463fc305337b02779</td><td> +<br/>www.semargroup.org, +<br/>www.ijsetr.com +<br/> +<br/>ISSN 2319-8885 +<br/>Vol.03,Issue.25 +<br/>September-2014, +<br/>Pages:5079-5085 +<br/>Design and Implementation of Robust Face Recognition System for +<br/>Uncontrolled Pose and Illumination Changes +<br/>2 +</td></tr><tr><td>f8ec92f6d009b588ddfbb47a518dd5e73855547d</td><td>J Inf Process Syst, Vol.10, No.3, pp.443~458, September 2014 +<br/> +<br/>ISSN 1976-913X (Print) +<br/>ISSN 2092-805X (Electronic) +<br/>Extreme Learning Machine Ensemble Using +<br/>Bagging for Facial Expression Recognition +</td></tr><tr><td>f869601ae682e6116daebefb77d92e7c5dd2cb15</td><td></td></tr><tr><td>f8ed5f2c71e1a647a82677df24e70cc46d2f12a8</td><td>International Journal of Scientific & Engineering Research, Volume 2, Issue 12, December-2011 1 +<br/>ISSN 2229-5518 +<br/>Artificial Neural Network Design and Parameter +<br/>Optimization for Facial Expressions Recognition +</td></tr><tr><td>cef841f27535c0865278ee9a4bc8ee113b4fb9f3</td><td></td></tr><tr><td>ce85d953086294d989c09ae5c41af795d098d5b2</td><td>This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. +<br/>Bilinear Analysis for Kernel Selection and +<br/>Nonlinear Feature Extraction +</td></tr><tr><td>ce691a37060944c136d2795e10ed7ba751cd8394</td><td></td></tr><tr><td>ce3f3088d0c0bf236638014a299a28e492069753</td><td></td></tr><tr><td>ce9a61bcba6decba72f91497085807bface02daf</td><td>Eigen-Harmonics Faces: Face Recognition under Generic Lighting +<br/>1Graduate School, CAS, Beijing, China, 100080 +<br/>2ICT-ISVISION Joint R&D Laboratory for Face Recognition, CAS, Beijing, China, 100080 +<br/>Emails: {lyqing, sgshan, wgao}jdl.ac.cn +</td></tr><tr><td>cef6cffd7ad15e7fa5632269ef154d32eaf057af</td><td>Emotion Detection Through Facial Feature +<br/>Recognition +<br/>through consistent +</td></tr><tr><td>cebfafea92ed51b74a8d27c730efdacd65572c40</td><td>JANUARY 2006 +<br/>31 +<br/>Matching 2.5D Face Scans to 3D Models +</td></tr><tr><td>ce54e891e956d5b502a834ad131616786897dc91</td><td>International Journal of Science and Research (IJSR) +<br/>ISSN (Online): 2319-7064 +<br/>Index Copernicus Value (2013): 6.14 | Impact Factor (2014): 5.611 +<br/>Face Recognition Using LTP Algorithm +<br/>1ECE & KUK +<br/>2Assistant Professor (ECE) +<br/>Volume 4 Issue 12, December 2015 +<br/>Licensed Under Creative Commons Attribution CC BY +<br/>www.ijsr.net +<br/> Variation in luminance: Third main challenge that +<br/>appears in face recognition process is the luminance. Due +<br/>to variation in the luminance the representation get varied +<br/>from the original image. The person with same poses +<br/>expression and seen from same viewpoint can be appear +<br/>very different due to variation in lightening. +</td></tr><tr><td>e0dedb6fc4d370f4399bf7d67e234dc44deb4333</td><td>Supplementary Material: Multi-Task Video Captioning with Video and +<br/>Entailment Generation +<br/>UNC Chapel Hill +<br/>1 Experimental Setup +<br/>1.1 Datasets +<br/>1.1.1 Video Captioning Datasets +<br/>YouTube2Text or MSVD The Microsoft Re- +<br/>search Video Description Corpus (MSVD) or +<br/>YouTube2Text (Chen and Dolan, 2011) is used +<br/>for our primary video captioning experiments. It +<br/>has 1970 YouTube videos in the wild with many +<br/>diverse captions in multiple languages for each +<br/>video. Caption annotations to these videos are +<br/>collected using Amazon Mechanical Turk (AMT). +<br/>All our experiments use only English captions. On +<br/>average, each video has 40 captions, and the over- +<br/>all dataset has about 80, 000 unique video-caption +<br/>pairs. The average clip duration is roughly 10 sec- +<br/>onds. We used the standard split as stated in Venu- +<br/>gopalan et al. (2015), i.e., 1200 videos for training, +<br/>100 videos for validation, and 670 for testing. +<br/>MSR-VTT MSR-VTT is a recent collection of +<br/>10, 000 video clips of 41.2 hours duration (i.e., +<br/>average duration of 15 seconds), which are an- +<br/>notated by AMT workers. It has 200, 000 video +<br/>clip-sentence pairs covering diverse content from +<br/>a commercial video search engine. On average, +<br/>each clip is annotated with 20 natural language +<br/>captions. We used the standard split as provided +<br/>in (Xu et al., 2016), i.e., 6, 513 video clips for +<br/>training, 497 for validation, and 2, 990 for testing. +<br/>M-VAD M-VAD is a movie description dataset +<br/>with 49, 000 video clips collected from 92 movies, +<br/>with the average clip duration being 6 seconds. +<br/>Alignment of descriptions to video clips is done +<br/>through an automatic procedure using Descrip- +<br/>tive Video Service (DVS) provided for the movies. +<br/>Each video clip description has only 1 or 2 sen- +<br/>tences, making most evaluation metrics (except +<br/>paraphrase-based METEOR) infeasible. Again, +<br/>we used the standard train/val/test split as pro- +<br/>vided in Torabi et al. (2015). +<br/>1.1.2 Video Prediction Dataset +<br/>For our unsupervised video representation learn- +<br/>ing task, we use the UCF-101 action videos +<br/>dataset (Soomro et al., 2012), which contains +<br/>13, 320 video clips of 101 action categories and +<br/>with an average clip length of 7.21 seconds each. +<br/>This dataset suits our video captioning task well +<br/>because both contain short video clips of a sin- +<br/>gle action or few actions, and hence using future +<br/>frame prediction on UCF-101 helps learn more ro- +<br/>bust and context-aware video representations for +<br/>our short clip video captioning task. We use the +<br/>standard split of 9, 500 videos for training (we +<br/>don’t need any validation set in our setup because +<br/>we directly tune on the validation set of the video +<br/>captioning task). +<br/>the +<br/>three +<br/>video +<br/>captioning +<br/>1.2 Pre-trained Visual Frame Features +<br/>For +<br/>datasets +<br/>(Youtube2Text, MSR-VTT, M-VAD) and the +<br/>unsupervised video prediction dataset (UCF-101), +<br/>we fix our sampling rate to 3f ps to bring uni- +<br/>formity in the temporal representation of actions +<br/>across all videos. These sampled frames are then +<br/>converted into features using several state-of-the- +<br/>art pre-trained models on ImageNet (Deng et al., +<br/>2009) – VGGNet +<br/>(Simonyan and Zisserman, +<br/>2015), GoogLeNet (Szegedy et al., 2015; Ioffe +<br/>and Szegedy, 2015), and Inception-v4 (Szegedy +<br/>et al., 2016). For VGGNet, we use its f c7 layer +<br/>features with dimension 4096. For GoogLeNet +<br/>and Inception-v4, we use the layer before the fully +<br/>connected layer with dimensions 1024 and 1536, +<br/>respectively. We follow standard preprocessing +<br/>and convert all the natural language descriptions +<br/>to lower case and tokenize the sentences and +<br/>remove punctuations. +</td></tr><tr><td>e096b11b3988441c0995c13742ad188a80f2b461</td><td>Noname manuscript No. +<br/>(will be inserted by the editor) +<br/>DeepProposals: Hunting Objects and Actions by Cascading +<br/>Deep Convolutional Layers +<br/>Van Gool +<br/>Received: date / Accepted: date +</td></tr><tr><td>e0c081a007435e0c64e208e9918ca727e2c1c44e</td><td></td></tr><tr><td>e00d4e4ba25fff3583b180db078ef962bf7d6824</td><td>Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 20 March 2017 doi:10.20944/preprints201703.0152.v1 +<br/>Article +<br/>Face Verification with Multi-Task and Multi-Scale +<br/>Features Fusion +</td></tr><tr><td>e0939b4518a5ad649ba04194f74f3413c793f28e</td><td>Technical Report +<br/>UCAM-CL-TR-636 +<br/>ISSN 1476-2986 +<br/>Number 636 +<br/>Computer Laboratory +<br/>Mind-reading machines: +<br/>automated inference +<br/>of complex mental states +<br/>July 2005 +<br/>15 JJ Thomson Avenue +<br/>Cambridge CB3 0FD +<br/>United Kingdom +<br/>phone +44 1223 763500 +<br/>http://www.cl.cam.ac.uk/ +</td></tr><tr><td>e0765de5cabe7e287582532456d7f4815acd74c1</td><td></td></tr><tr><td>e013c650c7c6b480a1b692bedb663947cd9d260f</td><td>860 +<br/>Robust Image Analysis With Sparse Representation +<br/>on Quantized Visual Features +</td></tr><tr><td>46a4551a6d53a3cd10474ef3945f546f45ef76ee</td><td>2014 IEEE Intelligent Vehicles Symposium (IV) +<br/>June 8-11, 2014. Dearborn, Michigan, USA +<br/>978-1-4799-3637-3/14/$31.00 ©2014 IEEE +<br/>344 +</td></tr><tr><td>4686bdcee01520ed6a769943f112b2471e436208</td><td>Utsumi et al. IPSJ Transactions on Computer Vision and +<br/>Applications (2017) 9:11 +<br/>DOI 10.1186/s41074-017-0024-5 +<br/>IPSJ Transactions on Computer +<br/>Vision and Applications +<br/>EXPRESS PAPER +<br/>Open Access +<br/>Fast search based on generalized +<br/>similarity measure +</td></tr><tr><td>4688787d064e59023a304f7c9af950d192ddd33e</td><td>Investigating the Discriminative Power of Keystroke +<br/>Sound +<br/>and Dimitris Metaxas, Member, IEEE +</td></tr><tr><td>46e86cdb674440f61b6658ef3e84fea95ea51fb4</td><td></td></tr><tr><td>464de30d3310123644ab81a1f0adc51598586fd2</td><td></td></tr><tr><td>466a5add15bb5f91e0cfd29a55f5fb159a7980e5</td><td>Video Repeat Recognition and Mining by Visual +<br/>Features +</td></tr><tr><td>46538b0d841654a0934e4c75ccd659f6c5309b72</td><td>Signal & Image Processing : An International Journal (SIPIJ) Vol.5, No.1, February 2014 +<br/>A NOVEL APPROACH TO GENERATE FACE +<br/>BIOMETRIC TEMPLATE USING BINARY +<br/>DISCRIMINATING ANALYSIS +<br/>1P.G. Student, Department of Computer Engineering, MCERC, Nashik (M.S.), India. +<br/>2Associate Professor, Department of Computer Engineering, +<br/>MCERC, Nashik (M.S.), India +</td></tr><tr><td>46196735a201185db3a6d8f6e473baf05ba7b68f</td><td></td></tr><tr><td>4682fee7dc045aea7177d7f3bfe344aabf153bd5</td><td>Tabula Rasa: Model Transfer for +<br/>Object Category Detection +<br/>Department of Engineering Science +<br/>Oxford +<br/>(Presented by Elad Liebman) +</td></tr><tr><td>2cbb4a2f8fd2ddac86f8804fd7ffacd830a66b58</td><td></td></tr><tr><td>2c8743089d9c7df04883405a31b5fbe494f175b4</td><td>Washington State Convention Center +<br/>Seattle, Washington, May 26-30, 2015 +<br/>978-1-4799-6922-7/15/$31.00 ©2015 IEEE +<br/>3039 +</td></tr><tr><td>2c61a9e26557dd0fe824909adeadf22a6a0d86b0</td><td></td></tr><tr><td>2c93c8da5dfe5c50119949881f90ac5a0a4f39fe</td><td>Advanced local motion patterns for macro and micro facial +<br/>expression recognition +<br/>B. Allaerta,∗, IM. Bilascoa, C. Djerabaa +<br/>aUniv. Lille, CNRS, Centrale Lille, UMR 9189 - CRIStAL - +<br/>Centre de Recherche en Informatique Signal et Automatique de Lille, F-59000 Lille, France +</td></tr><tr><td>2c2786ea6386f2d611fc9dbf209362699b104f83</td><td></td></tr><tr><td>2c848cc514293414d916c0e5931baf1e8583eabc</td><td>An automatic facial expression recognition system +<br/>evaluated by different classifiers +<br/>∗Programa de P´os-Graduac¸˜ao em Mecatrˆonica +<br/>Universidade Federal da Bahia, +<br/>†Department of Electrical Engineering - EESC/USP +</td></tr><tr><td>2cdd9e445e7259117b995516025fcfc02fa7eebb</td><td>Title +<br/>Temporal Exemplar-based Bayesian Networks for facial +<br/>expression recognition +<br/>Author(s) +<br/>Shang, L; Chan, KP +<br/>Citation +<br/>Proceedings - 7Th International Conference On Machine +<br/>Learning And Applications, Icmla 2008, 2008, p. 16-22 +<br/>Issued Date +<br/>2008 +<br/>URL +<br/>http://hdl.handle.net/10722/61208 +<br/>Rights +<br/>This work is licensed under a Creative Commons Attribution- +<br/>NonCommercial-NoDerivatives 4.0 International License.; +<br/>International Conference on Machine Learning and Applications +<br/>Proceedings. Copyright © IEEE.; ©2008 IEEE. Personal use of +<br/>this material is permitted. However, permission to +<br/>reprint/republish this material for advertising or promotional +<br/>purposes or for creating new collective works for resale or +<br/>redistribution to servers or lists, or to reuse any copyrighted +<br/>component of this work in other works must be obtained from +<br/>the IEEE. +</td></tr><tr><td>2c5d1e0719f3ad7f66e1763685ae536806f0c23b</td><td>AENet: Learning Deep Audio Features for Video +<br/>Analysis +</td></tr><tr><td>2c8f24f859bbbc4193d4d83645ef467bcf25adc2</td><td>845 +<br/>Classification in the Presence of +<br/>Label Noise: a Survey +</td></tr><tr><td>2cdde47c27a8ecd391cbb6b2dea64b73282c7491</td><td>ORDER-AWARE CONVOLUTIONAL POOLING FOR VIDEO BASED ACTION RECOGNITION +<br/>Order-aware Convolutional Pooling for Video Based +<br/>Action Recognition +</td></tr><tr><td>2c7c3a74da960cc76c00965bd3e343958464da45</td><td></td></tr><tr><td>2cf5f2091f9c2d9ab97086756c47cd11522a6ef3</td><td>MPIIGaze: Real-World Dataset and Deep +<br/>Appearance-Based Gaze Estimation +</td></tr><tr><td>79581c364cefe53bff6bdd224acd4f4bbc43d6d4</td><td></td></tr><tr><td>790aa543151312aef3f7102d64ea699a1d15cb29</td><td>Confidence-Weighted Local Expression Predictions for +<br/>Occlusion Handling in Expression Recognition and Action +<br/>Unit detection +<br/>1 Sorbonne Universités, UPMC Univ Paris 06, CNRS, ISIR UMR 7222 +<br/>4 place Jussieu 75005 Paris +</td></tr><tr><td>79f6a8f777a11fd626185ab549079236629431ac</td><td>Copyright +<br/>by +<br/>2013 +</td></tr><tr><td>795ea140df2c3d29753f40ccc4952ef24f46576c</td><td></td></tr><tr><td>79dc84a3bf76f1cb983902e2591d913cee5bdb0e</td><td></td></tr><tr><td>79b669abf65c2ca323098cf3f19fa7bdd837ff31</td><td> Deakin Research Online +<br/>This is the published version: +<br/>Rana, Santu, Liu, Wanquan, Lazarescu, Mihai and Venkatesh, Svetha 2008, Efficient tensor +<br/>based face recognition, in ICPR 2008 : Proceedings of the 19th International Conference on +<br/>Pattern Recognition, IEEE, Washington, D. C., pp. 1-4. +<br/>Available from Deakin Research Online: +<br/>http://hdl.handle.net/10536/DRO/DU:30044585 +<br/> +<br/>Reproduced with the kind permissions of the copyright owner. +<br/>Personal use of this material is permitted. However, permission to reprint/republish this +<br/>material for advertising or promotional purposes or for creating new collective works for +<br/>resale or redistribution to servers or lists, or to reuse any copyrighted component of this work +<br/>in other works must be obtained from the IEEE. +<br/>Copyright : 2008, IEEE +</td></tr><tr><td>79c3a7131c6c176b02b97d368cd0cd0bc713ff7e</td><td></td></tr><tr><td>79dd787b2877cf9ce08762d702589543bda373be</td><td>Face Detection Using SURF Cascade +<br/>Intel Labs China +</td></tr><tr><td>793e7f1ba18848908da30cbad14323b0389fd2a8</td><td></td></tr><tr><td>2dd6c988b279d89ab5fb5155baba65ce4ce53c1e</td><td></td></tr><tr><td>2d294c58b2afb529b26c49d3c92293431f5f98d0</td><td>4413 +<br/>Maximum Margin Projection Subspace Learning +<br/>for Visual Data Analysis +</td></tr><tr><td>2d1f86e2c7ba81392c8914edbc079ac64d29b666</td><td></td></tr><tr><td>2d05e768c64628c034db858b7154c6cbd580b2d5</td><td>Available Online at www.ijcsmc.com +<br/>International Journal of Computer Science and Mobile Computing +<br/> A Monthly Journal of Computer Science and Information Technology +<br/> IJCSMC, Vol. 4, Issue. 8, August 2015, pg.431 – 446 +<br/> RESEARCH ARTICLE +<br/>ISSN 2320–088X +<br/>FACIAL EXPRESSION RECOGNITION: +<br/>Machine Learning using C# +</td></tr><tr><td>2d072cd43de8d17ce3198fae4469c498f97c6277</td><td>Random Cascaded-Regression Copse for Robust +<br/>Facial Landmark Detection +<br/>and Xiao-Jun Wu +</td></tr><tr><td>2d71e0464a55ef2f424017ce91a6bcc6fd83f6c3</td><td>International Journal of Computer Applications (0975 – 8887) +<br/>National Conference on Advancements in Computer & Information Technology (NCACIT-2016) +<br/>A Survey on: Image Process using Two- Stage Crawler +<br/>Assistant Professor +<br/>SPPU, Pune +<br/>Department of Computer Engg +<br/>Department of Computer Engg +<br/>Department of Computer Engg +<br/>BE Student +<br/>SPPU, Pune +<br/>BE Student +<br/>SPPU, Pune +<br/>BE Student +<br/>Department of Computer Engg +<br/>SPPU, Pune +<br/>additional +<br/>analysis +<br/>for +<br/>information +</td></tr><tr><td>2d8d089d368f2982748fde93a959cf5944873673</td><td>Proceedings of NAACL-HLT 2018, pages 788–794 +<br/>New Orleans, Louisiana, June 1 - 6, 2018. c(cid:13)2018 Association for Computational Linguistics +<br/>788 +</td></tr><tr><td>2df4d05119fe3fbf1f8112b3ad901c33728b498a</td><td>Facial landmark detection using structured output deep +<br/>neural networks +<br/>Soufiane Belharbi ∗1, Cl´ement Chatelain∗1, Romain H´erault∗1, and S´ebastien +<br/>Adam∗2 +<br/>1LITIS EA 4108, INSA de Rouen, Saint ´Etienne du Rouvray 76800, France +<br/>2LITIS EA 4108, UFR des Sciences, Universit´e de Rouen, France. +<br/>September 24, 2015 +</td></tr><tr><td>4188bd3ef976ea0dec24a2512b44d7673fd4ad26</td><td>1050 +<br/>Nonlinear Non-Negative Component +<br/>Analysis Algorithms +</td></tr><tr><td>41000c3a3344676513ef4bfcd392d14c7a9a7599</td><td>A NOVEL APPROACH FOR GENERATING FACE +<br/>TEMPLATE USING BDA +<br/>1P.G. Student, Department of Computer Engineering, MCERC, Nashik (M.S.), India. +<br/>2Associate Professor, Department of Computer Engineering, MCERC, Nashik (M.S.), +<br/>India +</td></tr><tr><td>414715421e01e8c8b5743c5330e6d2553a08c16d</td><td>PoTion: Pose MoTion Representation for Action Recognition +<br/>1Inria∗ +<br/>2NAVER LABS Europe +</td></tr><tr><td>41ab4939db641fa4d327071ae9bb0df4a612dc89</td><td>Interpreting Face Images by Fitting a Fast +<br/>Illumination-Based 3D Active Appearance +<br/>Model +<br/>Instituto Nacional de Astrof´ısica, ´Optica y Electr´onica, +<br/>Luis Enrique Erro #1, 72840 Sta Ma. Tonantzintla. Pue., M´exico +<br/>Coordinaci´on de Ciencias Computacionales +</td></tr><tr><td>41a6196f88beced105d8bc48dd54d5494cc156fb</td><td>2015 International Conference on +<br/>Communications, Signal +<br/>Processing, and their Applications +<br/>(ICCSPA 2015) +<br/>Sharjah, United Arab Emirates +<br/>17-19 February 2015 +<br/>IEEE Catalog Number: +<br/>ISBN: +<br/>CFP1574T-POD +<br/>978-1-4799-6533-5 +</td></tr><tr><td>41de109bca9343691f1d5720df864cdbeeecd9d0</td><td>Article +<br/>Facial Emotion Recognition: A Survey and +<br/>Real-World User Experiences in Mixed Reality +<br/>Received: 10 December 2017; Accepted: 26 January 2018; Published: 1 Febuary 2018 +</td></tr><tr><td>41d9a240b711ff76c5448d4bf4df840cc5dad5fc</td><td>JOURNAL DRAFT, VOL. X, NO. X, APR 2013 +<br/>Image Similarity Using Sparse Representation +<br/>and Compression Distance +</td></tr><tr><td>419a6fca4c8d73a1e43003edc3f6b610174c41d2</td><td>A Component Based Approach Improves Classification of Discrete +<br/>Facial Expressions Over a Holistic Approach +</td></tr><tr><td>4180978dbcd09162d166f7449136cb0b320adf1f</td><td>Real-time head pose classification in uncontrolled environments +<br/>with Spatio-Temporal Active Appearance Models +<br/>∗ Matematica Aplicada i Analisi ,Universitat de Barcelona, Barcelona, Spain +<br/>+ Matematica Aplicada i Analisi, Universitat de Barcelona, Barcelona, Spain +<br/>+ Matematica Aplicada i Analisi, Universitat de Barcelona, Barcelona, Spain +</td></tr><tr><td>41b997f6cec7a6a773cd09f174cb6d2f036b36cd</td><td></td></tr><tr><td>413a184b584dc2b669fbe731ace1e48b22945443</td><td>Human Pose Co-Estimation and Applications +</td></tr><tr><td>83ca4cca9b28ae58f461b5a192e08dffdc1c76f3</td><td>DETECTING EMOTIONAL STRESS FROM FACIAL EXPRESSIONS FOR DRIVING SAFETY +<br/>Signal Processing Laboratory (LTS5), +<br/>´Ecole Polytechnique F´ed´erale de Lausanne, Switzerland +</td></tr><tr><td>831fbef657cc5e1bbf298ce6aad6b62f00a5b5d9</td><td></td></tr><tr><td>832e1d128059dd5ed5fa5a0b0f021a025903f9d5</td><td>Pairwise Conditional Random Forests for Facial Expression Recognition +<br/>S´everine Dubuisson1 +<br/>1 Sorbonne Universit´es, UPMC Univ Paris 06, CNRS, ISIR UMR 7222, 4 place Jussieu 75005 Paris +</td></tr><tr><td>83e093a07efcf795db5e3aa3576531d61557dd0d</td><td>Facial Landmark Localization using Robust +<br/>Relationship Priors and Approximative Gibbs +<br/>Sampling +<br/>Institut f¨ur Informationsverarbeitung (tnt) +<br/>Leibniz Universit¨at Hannover, Germany +</td></tr><tr><td>83b4899d2899dd6a8d956eda3c4b89f27f1cd308</td><td>1-4244-1437-7/07/$20.00 ©2007 IEEE +<br/>I - 377 +<br/>ICIP 2007 +</td></tr><tr><td>830e5b1043227fe189b3f93619ef4c58868758a7</td><td></td></tr><tr><td>8395cf3535a6628c3bdc9b8d0171568d551f5ff0</td><td>Entropy Non-increasing Games for the +<br/>Improvement of Dataflow Programming +<br/>Norbert B´atfai, Ren´at´o Besenczi, Gerg˝o Bogacsovics, +<br/>February 16, 2017 +</td></tr><tr><td>83ac942d71ba908c8d76fc68de6173151f012b38</td><td></td></tr><tr><td>834f5ab0cb374b13a6e19198d550e7a32901a4b2</td><td>Face Translation between Images and Videos using Identity-aware CycleGAN +<br/>†Computer Vision Lab, ETH Zurich, Switzerland +<br/>‡VISICS, KU Leuven, Belgium +</td></tr><tr><td>834b15762f97b4da11a2d851840123dbeee51d33</td><td>Landmark-free smile intensity estimation +<br/>IMAGO Research Group - Universidade Federal do Paran´a +<br/>Fig. 1. Overview of our method for smile intensity estimation +</td></tr><tr><td>833f6ab858f26b848f0d747de502127406f06417</td><td>978-1-4244-5654-3/09/$26.00 ©2009 IEEE +<br/>61 +<br/>ICIP 2009 +</td></tr><tr><td>8309e8f27f3fb6f2ac1b4343a4ad7db09fb8f0ff</td><td>Generic versus Salient Region-based Partitioning +<br/>for Local Appearance Face Recognition +<br/>Computer Science Depatment, Universit¨at Karlsruhe (TH) +<br/>Am Fasanengarten 5, Karlsruhe 76131, Germany +<br/>http://isl.ira.uka.de/cvhci +</td></tr><tr><td>1b55c4e804d1298cbbb9c507497177014a923d22</td><td>Incremental Class Representation +<br/>Learning for Face Recognition +<br/>Degree’s Thesis +<br/>Audiovisual Systems Engineering +<br/>Author: +<br/>Universitat Politècnica de Catalunya (UPC) +<br/>2016 - 2017 +</td></tr><tr><td>1bd50926079e68a6e32dc4412e9d5abe331daefb</td><td></td></tr><tr><td>1b150248d856f95da8316da868532a4286b9d58e</td><td>Analyzing 3D Objects in Cluttered Images +<br/>UC Irvine +<br/>UC Irvine +</td></tr><tr><td>1be498d4bbc30c3bfd0029114c784bc2114d67c0</td><td>Age and Gender Estimation of Unfiltered Faces +</td></tr><tr><td>1b300a7858ab7870d36622a51b0549b1936572d4</td><td>This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIP.2016.2537215, IEEE +<br/>Transactions on Image Processing +<br/>Dynamic Facial Expression Recognition with Atlas +<br/>Construction and Sparse Representation +</td></tr><tr><td>1b1173a3fb33f9dfaf8d8cc36eb0bf35e364913d</td><td>DICTA +<br/>#147 +<br/>000 +<br/>001 +<br/>002 +<br/>003 +<br/>004 +<br/>005 +<br/>006 +<br/>007 +<br/>008 +<br/>009 +<br/>010 +<br/>011 +<br/>012 +<br/>013 +<br/>014 +<br/>015 +<br/>016 +<br/>017 +<br/>018 +<br/>019 +<br/>020 +<br/>021 +<br/>022 +<br/>023 +<br/>024 +<br/>025 +<br/>026 +<br/>027 +<br/>028 +<br/>029 +<br/>030 +<br/>031 +<br/>032 +<br/>033 +<br/>034 +<br/>035 +<br/>036 +<br/>037 +<br/>038 +<br/>039 +<br/>040 +<br/>041 +<br/>042 +<br/>043 +<br/>044 +<br/>045 +<br/>046 +<br/>047 +<br/>048 +<br/>049 +<br/>050 +<br/>051 +<br/>052 +<br/>053 +<br/>DICTA 2010 Submission #147. CONFIDENTIAL REVIEW COPY. DO NOT DISTRIBUTE. +<br/>Registration Invariant Representations for Expression Detection +<br/>Anonymous DICTA submission +<br/>Paper ID 147 +</td></tr><tr><td>1b0a071450c419138432c033f722027ec88846ea</td><td>Windsor Oceanico Hotel, Rio de Janeiro, Brazil, November 1-4, 2016 +<br/>978-1-5090-1889-5/16/$31.00 ©2016 IEEE +<br/>649 +</td></tr><tr><td>1b3b01513f99d13973e631c87ffa43904cd8a821</td><td>HMM RECOGNITION OF EXPRESSIONS IN UNRESTRAINED VIDEO INTERVALS +<br/>Universitat Politècnica de Catalunya, Barcelona, Spain +</td></tr><tr><td>1bc214c39536c940b12c3a2a6b78cafcbfddb59a</td><td></td></tr><tr><td>1be18a701d5af2d8088db3e6aaa5b9b1d54b6fd3</td><td>ENHANCEMENT OF FAST FACE DETECTION ALGORITHM BASED ON A CASCADE OF +<br/>DECISION TREES +<br/>Commission II, WG II/5 +<br/>KEY WORDS: Face Detection, Cascade Algorithm, Decision Trees. +</td></tr><tr><td>1b79628af96eb3ad64dbb859dae64f31a09027d5</td><td></td></tr><tr><td>1b4f6f73c70353869026e5eec1dd903f9e26d43f</td><td>Robust Subjective Visual Property Prediction +<br/>from Crowdsourced Pairwise Labels +</td></tr><tr><td>1bc23c771688109bed9fd295ce82d7e702726327</td><td></td></tr><tr><td>1b589016fbabe607a1fb7ce0c265442be9caf3a9</td><td></td></tr><tr><td>1b27ca161d2e1d4dd7d22b1247acee5c53db5104</td><td></td></tr><tr><td>7711a7404f1f1ac3a0107203936e6332f50ac30c</td><td>Action Classification and Highlighting in Videos +<br/>Disney Research Pittsburgh +<br/>Disney Research Pittsburgh +</td></tr><tr><td>778c9f88839eb26129427e1b8633caa4bd4d275e</td><td>Pose Pooling Kernels for Sub-category Recognition +<br/>ICSI & UC Berkeley +<br/>ICSI & UC Berkeley +<br/>Trever Darrell +<br/>ICSI & UC Berkeley +</td></tr><tr><td>7789a5d87884f8bafec8a82085292e87d4e2866f</td><td>A Unified Tensor-based Active Appearance Face +<br/>Model +<br/>Member, IEEE +</td></tr><tr><td>776835eb176ed4655d6e6c308ab203126194c41e</td><td></td></tr><tr><td>778bff335ae1b77fd7ec67404f71a1446624331b</td><td>Hough Forest-based Facial Expression Recognition from +<br/>Video Sequences +<br/>BIWI, ETH Zurich http://www.vision.ee.ethz.ch +<br/>VISICS, K.U. Leuven http://www.esat.kuleuven.be/psi/visics +</td></tr><tr><td>7726a6ab26a1654d34ec04c0b7b3dd80c5f84e0d</td><td>CONTENT-AWARE COMPRESSION USING SALIENCY-DRIVEN IMAGE RETARGETING +<br/>*Disney Research Zurich +<br/>†ETH Zurich +</td></tr><tr><td>7754b708d6258fb8279aa5667ce805e9f925dfd0</td><td>Facial Action Unit Recognition by Exploiting +<br/>Their Dynamic and Semantic Relationships +</td></tr><tr><td>77db171a523fc3d08c91cea94c9562f3edce56e1</td><td>Poursaberi et al. EURASIP Journal on Image and Video Processing 2012, 2012:17 +<br/>http://jivp.eurasipjournals.com/content/2012/1/17 +<br/>R ES EAR CH +<br/>Open Access +<br/>Gauss–Laguerre wavelet textural feature fusion +<br/>with geometrical information for facial expression +<br/>identification +</td></tr><tr><td>77037a22c9b8169930d74d2ce6f50f1a999c1221</td><td>Robust Face Recognition With Kernelized +<br/>Locality-Sensitive Group Sparsity Representation +</td></tr><tr><td>77d31d2ec25df44781d999d6ff980183093fb3de</td><td>The Multiverse Loss for Robust Transfer Learning +<br/>Supplementary +<br/>1. Omitted proofs +<br/>for which the joint loss: +<br/>m(cid:88) +<br/>r=1 +<br/>L(F r, br, D, y) +<br/>(2) +<br/>J(F 1, b1...F m, bm, D, y) = +<br/>is bounded by: +<br/>mL∗(D, y) ≤ J(F 1, b1...F m, bm, D, y) +<br/>m−1(cid:88) +<br/>≤ mL∗(D, y) + +<br/>Alλd−j+1 +<br/>(3) +<br/>l=1 +<br/>where [A1 . . . Am−1] are bounded parameters. +<br/>We provide proofs that were omitted from the paper for +<br/>lack of space. We follow the same theorem numbering as in +<br/>the paper. +<br/>Lemma 1. The minimizers F ∗, b∗ of L are not unique, and +<br/>it holds that for any vector v ∈ Rc and scalar s, the solu- +<br/>tions F ∗ + v1(cid:62) +<br/>Proof. denoting V = v1(cid:62) +<br/>c , b∗ + s1c are also minimizers of L. +<br/>c , s = s1c, +<br/>i v+byi +s +<br/>i v+bj +s +<br/>i fyi +byi +<br/>i v+sed(cid:62) +<br/>i fj +bj +<br/>i=1 +<br/>log( +<br/>L(F ∗ + V, b∗ + s, D, y) = +<br/>i fyi +d(cid:62) +<br/>ed(cid:62) +<br/>i fj +d(cid:62) +<br/>j=1 ed(cid:62) +<br/>i v+sed(cid:62) +<br/>ed(cid:62) +<br/>j=1 ed(cid:62) +<br/>i v+sed(cid:62) +<br/>ed(cid:62) +<br/>(cid:80)c +<br/>(cid:80)c +<br/>i v+s(cid:80)c +<br/>− n(cid:88) +<br/>= − n(cid:88) +<br/>= − n(cid:88) +<br/>(cid:80)c +<br/>= − n(cid:88) +<br/>ed(cid:62) +<br/>i fyi +byi +<br/>j=1 ed(cid:62) +<br/>i fj +bj +<br/>ed(cid:62) +<br/>log( +<br/>log( +<br/>log( +<br/>i=1 +<br/>i=1 +<br/>i=1 +<br/>i fj +bj +<br/>i fyi +byi +<br/>j=1 ed(cid:62) +<br/>) = L(F ∗, b∗, D, y) +<br/>The following simple lemma was not part of the paper. +<br/>However, it is the reasoning behind the statement at the end +<br/>of the proof of Thm. 1. “Since ∀i, j pi(j) > 0 and since +<br/>rank(D) is full,(cid:80)n +<br/>Lemma 2. Let K =(cid:80)n +<br/>such that ∀i qi > 0, the matrix ˆK =(cid:80)n +<br/>i be a full rank d×d matrix, +<br/>i.e., it is PD and not just PSD, then for all vector q ∈ Rn +<br/>is also +<br/>i pi(j)pi(j(cid:48)) is PD.” +<br/>i=1 did(cid:62) +<br/>i=1 did(cid:62) +<br/>i=1 qidid(cid:62) +<br/>full rank. +<br/>Proof. For +<br/>(miniqi)v(cid:62)Kv > 0. +<br/>every vector v +<br/>(cid:2)f 1 +<br/>(cid:3) , b1, F 2 = (cid:2)f 2 +<br/>Theorem 3. There exist a set of weights F 1 = +<br/>j ⊥ f s +<br/>C ] , bm which are orthogonal ∀jrs f r +<br/>2 , ..., f 1 +<br/>2 , ..., f m +<br/>1 , f 1 +<br/>1 , f m +<br/>2 , ..., f 2 +<br/>1 , f 2 +<br/>[f m +<br/>(cid:3) , b2...F m = +<br/>Proof. We again prove the theorem by constructing such a +<br/>solution. Denoting by vd−m+2...vd the eigenvectors of K +<br/>corresponding to λd−m+2 . . . λd. Given F 1 = F ∗, b1 = b∗, +<br/>we can construct each pair F r, br as follows: +<br/>(1) +<br/>∀j, r +<br/>fj +<br/>r = f1 +<br/>1 + +<br/>m−1(cid:88) +<br/>l=1 +<br/>αjlrvd−l+1 +<br/>br = b1 +<br/>(4) +<br/>The tensor of parameters αjlr is constructed to insure the +<br/>orthogonality condition. Formally, αjlr has to satisfy: +<br/>Rd, +<br/>v(cid:62) ˆKv +<br/>∀j, r (cid:54)= s +<br/>(f 1 +<br/>j + +<br/>m−1(cid:88) +<br/>l=1 +<br/>αjlrvd−l+1)(cid:62)f s +<br/>j = 0 +<br/>(5) +<br/>2 m(m− 1) equations, it +<br/>Noticing that 5 constitutes a set of 1 +<br/>can be satisfied by the tensor αjlr which contains m(m − +<br/>c ] = F r − +<br/>1)c parameters. Defining Ψr = [ψr +<br/>1, ψr +<br/>2, . . . , ψr +</td></tr><tr><td>486840f4f524e97f692a7f6b42cd19019ee71533</td><td>DeepVisage: Making face recognition simple yet with powerful generalization +<br/>skills +<br/>1Laboratoire LIRIS, ´Ecole centrale de Lyon, 69134 Ecully, France. +<br/>2Safran Identity & Security, 92130 Issy-les-Moulineaux, France. +</td></tr><tr><td>48186494fc7c0cc664edec16ce582b3fcb5249c0</td><td>P-CNN: Pose-based CNN Features for Action Recognition +<br/>Guilhem Ch´eron∗ † +<br/>INRIA +</td></tr><tr><td>48499deeaa1e31ac22c901d115b8b9867f89f952</td><td>Interim Report of Final Year Project +<br/>HKU-Face: A Large Scale Dataset for +<br/>Deep Face Recognition +<br/>3035140108 +<br/>Haoyu Li +<br/>3035141841 +<br/>COMP4801 Final Year Project +<br/>Project Code: 17007 +</td></tr><tr><td>486a82f50835ea888fbc5c6babf3cf8e8b9807bc</td><td>MSU TECHNICAL REPORT MSU-CSE-15-11, JULY 24, 2015 +<br/>Face Search at Scale: 80 Million Gallery +</td></tr><tr><td>4866a5d6d7a40a26f038fc743e16345c064e9842</td><td></td></tr><tr><td>487df616e981557c8e1201829a1d0ec1ecb7d275</td><td>Acoustic Echo Cancellation Using a Vector-Space-Based +<br/>Adaptive Filtering Algorithm +</td></tr><tr><td>48f211a9764f2bf6d6dda4a467008eda5680837a</td><td></td></tr><tr><td>4858d014bb5119a199448fcd36746c413e60f295</td><td></td></tr><tr><td>48cfc5789c246c6ad88ff841701204fc9d6577ed</td><td>J Inf Process Syst, Vol.12, No.3, pp.392~409, September 2016 +<br/> +<br/> +<br/>ISSN 1976-913X (Print) +<br/>ISSN 2092-805X (Electronic) +<br/>Age Invariant Face Recognition Based on DCT +<br/>Feature Extraction and Kernel Fisher Analysis +</td></tr><tr><td>70f189798c8b9f2b31c8b5566a5cf3107050b349</td><td>The Challenge of Face Recognition from Digital Point-and-Shoot Cameras +<br/>David Bolme‡ +</td></tr><tr><td>70109c670471db2e0ede3842cbb58ba6be804561</td><td>Noname manuscript No. +<br/>(will be inserted by the editor) +<br/>Zero-Shot Visual Recognition via Bidirectional Latent Embedding +<br/>Received: date / Accepted: date +</td></tr><tr><td>703890b7a50d6535900a5883e8d2a6813ead3a03</td><td></td></tr><tr><td>706236308e1c8d8b8ba7749869c6b9c25fa9f957</td><td>Crowdsourced Data Collection of Facial Responses +<br/>MIT Media Lab +<br/>Cambridge +<br/>02139, USA +<br/>Rosalind Picard +<br/>MIT Media Lab +<br/>Cambridge +<br/>02139, USA +<br/>MIT Media Lab +<br/>Cambridge +<br/>02139, USA +</td></tr><tr><td>70569810e46f476515fce80a602a210f8d9a2b95</td><td>Apparent Age Estimation from Face Images Combining General and +<br/>Children-Specialized Deep Learning Models +<br/>1Orange Labs – France Telecom, 4 rue Clos Courtel, 35512 Cesson-S´evign´e, France +<br/>2Eurecom, 450 route des Chappes, 06410 Biot, France +</td></tr><tr><td>70e79d7b64f5540d309465620b0dab19d9520df1</td><td>International Journal of Scientific & Engineering Research, Volume 8, Issue 3, March-2017 +<br/>ISSN 2229-5518 +<br/>Facial Expression Recognition System +<br/>Using Extreme Learning Machine +</td></tr><tr><td>7003d903d5e88351d649b90d378f3fc5f211282b</td><td>International Journal of Computer Applications (0975 – 8887) +<br/>Volume 68– No.23, April 2013 +<br/>Facial Expression Recognition using Gabor Wavelet +<br/>ENTC SVERI’S COE (Poly), +<br/>Pandharpur, +<br/>Solapur, India +<br/>ENTC SVERI’S COE, +<br/>Pandharpur, +<br/>Solapur, India +<br/>ENTC SVERI’S COE (Poly), +<br/>Pandharpur, +<br/>Solapur, India +</td></tr><tr><td>70bf1769d2d5737fc82de72c24adbb7882d2effd</td><td>Face detection in intelligent ambiences with colored illumination +<br/>Department of Intelligent Systems +<br/>TU Delft +<br/>Delft, The Netherlands +</td></tr><tr><td>1e799047e294267087ec1e2c385fac67074ee5c8</td><td>IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 21, NO. 12, DECEMBER 1999 +<br/>1357 +<br/>Short Papers___________________________________________________________________________________________________ +<br/>Automatic Classification of +<br/>Single Facial Images +</td></tr><tr><td>1ef4815f41fa3a9217a8a8af12cc385f6ed137e1</td><td>Rendering of Eyes for Eye-Shape Registration and Gaze Estimation +</td></tr><tr><td>1e7ae86a78a9b4860aa720fb0fd0bdc199b092c3</td><td>Article +<br/>A Brief Review of Facial Emotion Recognition Based +<br/>on Visual Information +<br/>Byoung Chul Ko ID +<br/>Tel.: +82-10-3559-4564 +<br/>Received: 6 December 2017; Accepted: 25 January 2018; Published: 30 January 2018 +</td></tr><tr><td>1e8eee51fd3bf7a9570d6ee6aa9a09454254689d</td><td>This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPAMI.2016.2582166, IEEE +<br/>Transactions on Pattern Analysis and Machine Intelligence +<br/>Face Search at Scale +</td></tr><tr><td>1ea8085fe1c79d12adffb02bd157b54d799568e4</td><td></td></tr><tr><td>1ebdfceebad642299e573a8995bc5ed1fad173e3</td><td></td></tr><tr><td>1eec03527703114d15e98ef9e55bee5d6eeba736</td><td>UNIVERSITÄT KARLSRUHE (TH) +<br/>FAKULTÄT FÜR INFORMATIK +<br/>INTERACTIVE SYSTEMS LABS +<br/>DIPLOMA THESIS +<br/>Automatic identification +<br/>of persons in TV series +<br/>SUBMITTED BY +<br/>MAY 2008 +<br/>ADVISORS +</td></tr><tr><td>1e8394cc9fe7c2392aa36fb4878faf7e78bbf2de</td><td>TO APPEAR IN IEEE THMS +<br/>Zero-Shot Object Recognition System +<br/>based on Topic Model +</td></tr><tr><td>1ef4aac0ebc34e76123f848c256840d89ff728d0</td><td></td></tr><tr><td>1ecb56e7c06a380b3ce582af3a629f6ef0104457</td><td>List of Contents Vol.8 +<br/>Contents of +<br/>Journal of Advanced Computational +<br/> Intelligence and Intelligent Informatics +<br/>Volume 8 +<br/>Vol.8 No.1, January 2004 +<br/>Editorial: +<br/>o Special Issue on Selected Papers from Humanoid, +<br/>Papers: +<br/>o Dynamic Color Object Recognition Using Fuzzy +<br/>Nano-technology, Information Technology, +<br/>Communication and Control, Environment, and +<br/>Management (HNICEM’03). +<br/>. 1 +<br/>Elmer P. Dadios +<br/>Papers: +<br/>o A New Way of Discovery of Belief, Desire and +<br/>Intention in the BDI Agent-Based Software +<br/>Modeling . +<br/>. 2 +<br/>o Integration of Distributed Robotic Systems +<br/>. 7 +<br/>Fakhri Karray, Rogelio Soto, Federico Guedea, +<br/>and Insop Song +<br/>o A Searching and Tracking Framework for +<br/>Multi-Robot Observation of Multiple Moving +<br/>Targets . +<br/>. 14 +<br/>Zheng Liu, Marcelo H. Ang Jr., and Winston +<br/>Khoon Guan Seah +<br/>Development Paper: +<br/>o Possibilistic Uncertainty Propagation and +<br/>Compromise Programming in the Life Cycle +<br/>Analysis of Alternative Motor Vehicle Fuels +<br/>Raymond R. Tan, Alvin B. Culaba, and +<br/>Michael R. I. Purvis +<br/>. 23 +<br/>Logic . +<br/>Napoleon H. Reyes, and Elmer P. Dadios +<br/>. 29 +<br/>o A Optical Coordinate Measuring Machine for +<br/>Nanoscale Dimensional Metrology . +<br/>. 39 +<br/>Eric Kirkland, Thomas R. Kurfess, and Steven +<br/>Y. Liang +<br/>o Humanoid Robot HanSaRam: Recent Progress +<br/>and Developments . +<br/>. 45 +<br/>Jong-Hwan Kim, Dong-Han Kim, Yong-Jae +<br/>Kim, Kui-Hong Park, Jae-Ho Park, +<br/>Choon-Kyoung Moon, Jee-Hwan Ryu, Kiam +<br/>Tian Seow, and Kyoung-Chul Koh +<br/>o Generalized Associative Memory Models: Their +<br/>Memory Capacities and Potential Application +<br/>. 56 +<br/>Teddy N. Yap, Jr., and Arnulfo P. Azcarraga +<br/>o Hybrid Fuzzy Logic Strategy for Soccer Robot +<br/>Game. +<br/>. 65 +<br/>Elmer A. Maravillas , Napoleon H. Reyes, and +<br/>Elmer P. Dadios +<br/>o Image Compression and Reconstruction Based on +<br/>Fuzzy Relation and Soft Computing +<br/>Technology . +<br/>. 72 +<br/>Kaoru Hirota, Hajime Nobuhara, Kazuhiko +<br/>Kawamoto, and Shin’ichi Yoshida +<br/>Vol.8 No.2, March 2004 +<br/>Editorial: +<br/>o Special Issue on Pattern Recognition . +<br/>. 83 +<br/>Papers: +<br/>o Operation of Spatiotemporal Patterns Stored in +<br/>Osamu Hasegawa +<br/>Review: +<br/>o Support Vector Machine and Generalization . 84 +<br/>Takio Kurita +<br/>o Bayesian Network: Probabilistic Reasoning, +<br/>Statistical Learning, and Applications . +<br/>. 93 +<br/>Yoichi Motomura +<br/>Living Neuronal Networks Cultured on a +<br/>Microelectrode Array . +<br/>Suguru N. Kudoh, and Takahisa Taguchi +<br/>o Rapid Discriminative Learning . +<br/>. 100 +<br/>. 108 +<br/>Jun Rokui +<br/>o Robust Fuzzy Clustering Based on Similarity +<br/>between Data . +<br/>Kohei Inoue, and Kiichi Urahama +<br/>Vol.8 No.6, 2004 +<br/>Journal of Advanced Computational Intelligence +<br/>and Intelligent Informatics +<br/>. 115 +<br/>I-1 +</td></tr><tr><td>1e64b2d2f0a8a608d0d9d913c4baee6973995952</td><td>DOMINANT AND +<br/>COMPLEMENTARY MULTI- +<br/>EMOTIONAL FACIAL +<br/>EXPRESSION RECOGNITION +<br/>USING C-SUPPORT VECTOR +<br/>CLASSIFICATION +</td></tr><tr><td>1e21b925b65303ef0299af65e018ec1e1b9b8d60</td><td>Under review as a conference paper at ICLR 2017 +<br/>UNSUPERVISED CROSS-DOMAIN IMAGE GENERATION +<br/>Facebook AI Research +<br/>Tel-Aviv, Israel +</td></tr><tr><td>1ee27c66fabde8ffe90bd2f4ccee5835f8dedbb9</td><td>Entropy Regularization +<br/>The problem of semi-supervised induction consists in learning a decision rule from +<br/>labeled and unlabeled data. This task can be undertaken by discriminative methods, +<br/>provided that learning criteria are adapted consequently. In this chapter, we moti- +<br/>vate the use of entropy regularization as a means to bene(cid:12)t from unlabeled data in +<br/>the framework of maximum a posteriori estimation. The learning criterion is derived +<br/>from clearly stated assumptions and can be applied to any smoothly parametrized +<br/>model of posterior probabilities. The regularization scheme favors low density sep- +<br/>aration, without any modeling of the density of input features. The contribution +<br/>of unlabeled data to the learning criterion induces local optima, but this problem +<br/>can be alleviated by deterministic annealing. For well-behaved models of posterior +<br/>probabilities, deterministic annealing EM provides a decomposition of the learning +<br/>problem in a series of concave subproblems. Other approaches to the semi-supervised +<br/>problem are shown to be close relatives or limiting cases of entropy regularization. +<br/>A series of experiments illustrates the good behavior of the algorithm in terms of +<br/>performance and robustness with respect to the violation of the postulated low den- +<br/>sity separation assumption. The minimum entropy solution bene(cid:12)ts from unlabeled +<br/>data and is able to challenge mixture models and manifold learning in a number of +<br/>situations. +<br/>9.1 Introduction +<br/>semi-supervised +<br/>induction +<br/>This chapter addresses semi-supervised induction, which refers to the learning of +<br/>a decision rule, on the entire input domain X, from labeled and unlabeled data. +<br/>The objective is identical to the one of supervised classi(cid:12)cation: generalize from +<br/>examples. The problem di(cid:11)ers in the respect that the supervisor’s responses are +<br/>missing for some training examples. This characteristic is shared with transduction, +<br/>which has however a di(cid:11)erent goal, that is, of predicting labels on a set of prede(cid:12)ned +</td></tr><tr><td>1ee3b4ba04e54bfbacba94d54bf8d05fd202931d</td><td>Indonesian Journal of Electrical Engineering and Computer Science +<br/>Vol. 12, No. 2, November 2018, pp. 476~481 +<br/>ISSN: 2502-4752, DOI: 10.11591/ijeecs.v12.i2.pp476-481 +<br/> 476 +<br/>Celebrity Face Recognition using Deep Learning +<br/>1,2,3Faculty of Computer and Mathematical Sciences, UniversitiTeknologi MARA (UiTM), +<br/>4Faculty of Computer and Mathematical Sciences, UniversitiTeknologi MARA (UiTM), +<br/> Shah Alam, Selangor, Malaysia +<br/>Campus Jasin, Melaka, Malaysia +<br/>Article Info +<br/>Article history: +<br/>Received May 29, 2018 +<br/>Revised Jul 30, 2018 +<br/>Accepted Aug 3, 2018 +<br/>Keywords: +<br/>AlexNet +<br/>Convolutional neural network +<br/>Deep learning +<br/>Face recognition +<br/>GoogLeNet +</td></tr><tr><td>1e41a3fdaac9f306c0ef0a978ae050d884d77d2a</td><td>411 +<br/>Robust Object Recognition with +<br/>Cortex-Like Mechanisms +<br/>Tomaso Poggio, Member, IEEE +</td></tr><tr><td>1e1e66783f51a206509b0a427e68b3f6e40a27c8</td><td>SEMI-SUPERVISED ESTIMATION OF PERCEIVED AGE +<br/>FROM FACE IMAGES +<br/>VALWAY Technology Center, NEC Soft, Ltd., Tokyo, Japan +<br/>Keywords: +</td></tr><tr><td>1efaa128378f988965841eb3f49d1319a102dc36</td><td>JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015 +<br/>Hierarchical binary CNNs for landmark +<br/>localization with limited resources +</td></tr><tr><td>8451bf3dd6bcd946be14b1a75af8bbb65a42d4b2</td><td>Consensual and Privacy-Preserving Sharing of +<br/>Multi-Subject and Interdependent Data +<br/>EPFL, UNIL–HEC Lausanne +<br/>K´evin Huguenin +<br/>UNIL–HEC Lausanne +<br/>EPFL +<br/>EPFL +</td></tr><tr><td>84fe5b4ac805af63206012d29523a1e033bc827e</td><td></td></tr><tr><td>84e4b7469f9c4b6c9e73733fa28788730fd30379</td><td>Duong et al. EURASIP Journal on Advances in Signal Processing (2018) 2018:10 +<br/>DOI 10.1186/s13634-017-0521-9 +<br/>EURASIP Journal on Advances +<br/>in Signal Processing +<br/>R ES EAR CH +<br/>Projective complex matrix factorization for +<br/>facial expression recognition +<br/>Open Access +</td></tr><tr><td>84dcf04802743d9907b5b3ae28b19cbbacd97981</td><td></td></tr><tr><td>84fa126cb19d569d2f0147bf6f9e26b54c9ad4f1</td><td>Improved Boosting Performance by Explicit +<br/>Handling of Ambiguous Positive Examples +</td></tr><tr><td>841a5de1d71a0b51957d9be9d9bebed33fb5d9fa</td><td>5017 +<br/>PCANet: A Simple Deep Learning Baseline for +<br/>Image Classification? +</td></tr><tr><td>849f891973ad2b6c6f70d7d43d9ac5805f1a1a5b</td><td>Detecting Faces Using Region-based Fully +<br/>Convolutional Networks +<br/>Tencent AI Lab, China +</td></tr><tr><td>4adca62f888226d3a16654ca499bf2a7d3d11b71</td><td>Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics, pages 572–582, +<br/>Sofia, Bulgaria, August 4-9 2013. c(cid:13)2013 Association for Computational Linguistics +<br/>572 +</td></tr><tr><td>4a2d54ea1da851151d43b38652b7ea30cdb6dfb2</td><td>Direct Recognition of Motion Blurred Faces +</td></tr><tr><td>4a3758f283b7c484d3f164528d73bc8667eb1591</td><td>Attribute Enhanced Face Aging with Wavelet-based Generative Adversarial +<br/>Networks +<br/>Center for Research on Intelligent Perception and Computing, CASIA +<br/>National Laboratory of Pattern Recognition, CASIA +</td></tr><tr><td>4abd49538d04ea5c7e6d31701b57ea17bc349412</td><td>Recognizing Fine-Grained and Composite Activities +<br/>using Hand-Centric Features and Script Data +</td></tr><tr><td>4a0f98d7dbc31497106d4f652968c708f7da6692</td><td>Real-time Eye Gaze Direction Classification Using +<br/>Convolutional Neural Network +</td></tr><tr><td>4acd683b5f91589002e6f50885df51f48bc985f4</td><td>BRIDGING COMPUTER VISION AND SOCIAL SCIENCE : A MULTI-CAMERA VISION +<br/>SYSTEM FOR SOCIAL INTERACTION TRAINING ANALYSIS +<br/>Peter Tu +<br/>GE Global Research, Niskayuna NY USA +</td></tr><tr><td>4aeb87c11fb3a8ad603311c4650040fd3c088832</td><td>Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence (IJCAI-17) +<br/>1816 +</td></tr><tr><td>4a3d96b2a53114da4be3880f652a6eef3f3cc035</td><td>2666 +<br/>A Dictionary Learning-Based +<br/>3D Morphable Shape Model +</td></tr><tr><td>4a6fcf714f663618657effc341ae5961784504c7</td><td>Scaling up Class-Specific Kernel Discriminant +<br/>Analysis for large-scale Face Verification +</td></tr><tr><td>24115d209e0733e319e39badc5411bbfd82c5133</td><td>Long-term Recurrent Convolutional Networks for +<br/>Visual Recognition and Description +</td></tr><tr><td>24c442ac3f6802296d71b1a1914b5d44e48b4f29</td><td>Pose and expression-coherent face recovery in the wild +<br/>Technicolor, Cesson-S´evign´e, France +<br/>Franc¸ois Le Clerc +<br/>Patrick P´erez +</td></tr><tr><td>24aac045f1e1a4c13a58eab4c7618dccd4c0e671</td><td></td></tr><tr><td>240d5390af19bb43761f112b0209771f19bfb696</td><td></td></tr><tr><td>24e099e77ae7bae3df2bebdc0ee4e00acca71250</td><td>Robust face alignment under occlusion via regional predictive power +<br/>estimation. +<br/>© 2015 IEEE +<br/>For additional information about this publication click this link. +<br/>http://qmro.qmul.ac.uk/xmlui/handle/123456789/22467 +<br/>Information about this research object was correct at the time of download; we occasionally +<br/>make corrections to records, please therefore check the published record when citing. For +</td></tr><tr><td>2450c618cca4cbd9b8cdbdb05bb57d67e63069b1</td><td>A Connexionist Approach for Robust and Precise Facial Feature Detection in +<br/>Complex Scenes +<br/>Stefan Duffner and Christophe Garcia +<br/>France Telecom Research & Development +<br/>4, rue du Clos Courtel +<br/>35512 Cesson-S´evign´e, France +</td></tr><tr><td>244b57cc4a00076efd5f913cc2833138087e1258</td><td>Warped Convolutions: Efficient Invariance to Spatial Transformations +</td></tr><tr><td>24869258fef8f47623b5ef43bd978a525f0af60e</td><td><b>UNIVERSITÉDEGRENOBLENoattribuéparlabibliothèqueTHÈSEpourobtenirlegradedeDOCTEURDEL’UNIVERSITÉDEGRENOBLESpécialité:MathématiquesetInformatiquepréparéeauLaboratoireJeanKuntzmanndanslecadredel’ÉcoleDoctoraleMathématiques,SciencesetTechnologiesdel’Information,InformatiqueprésentéeetsoutenuepubliquementparMatthieuGuillauminle27septembre2010ExploitingMultimodalDataforImageUnderstandingDonnéesmultimodalespourl’analysed’imageDirecteursdethèse:CordeliaSchmidetJakobVerbeekJURYM.ÉricGaussierUniversitéJosephFourierPrésidentM.AntonioTorralbaMassachusettsInstituteofTechnologyRapporteurMmeTinneTuytelaarsKatholiekeUniversiteitLeuvenRapporteurM.MarkEveringhamUniversityofLeedsExaminateurMmeCordeliaSchmidINRIAGrenobleExaminatriceM.JakobVerbeekINRIAGrenobleExaminateur</b></td></tr><tr><td>24d376e4d580fb28fd66bc5e7681f1a8db3b6b78</td><td></td></tr><tr><td>24ff832171cb774087a614152c21f54589bf7523</td><td>Beat-Event Detection in Action Movie Franchises +<br/>Jerome Revaud +<br/>Zaid Harchaoui +</td></tr><tr><td>24bf94f8090daf9bda56d54e42009067839b20df</td><td></td></tr><tr><td>230527d37421c28b7387c54e203deda64564e1b7</td><td>Person Re-identification: System Design and +<br/>Evaluation Overview +</td></tr><tr><td>23fdbef123bcda0f07d940c72f3b15704fd49a98</td><td></td></tr><tr><td>23ebbbba11c6ca785b0589543bf5675883283a57</td><td></td></tr><tr><td>23172f9a397f13ae1ecb5793efd81b6aba9b4537</td><td>Proceedings of the 2015 Workshop on Vision and Language (VL’15), pages 10–17, +<br/>Lisbon, Portugal, 18 September 2015. c(cid:13)2015 Association for Computational Linguistics. +<br/>10 +</td></tr><tr><td>236a4f38f79a4dcc2183e99b568f472cf45d27f4</td><td>1632 +<br/>Randomized Clustering Forests +<br/>for Image Classification +<br/>Frederic Jurie, Member, IEEE Computer Society +</td></tr><tr><td>230c4a30f439700355b268e5f57d15851bcbf41f</td><td>EM Algorithms for Weighted-Data Clustering +<br/>with Application to Audio-Visual Scene Analysis +</td></tr><tr><td>237fa91c8e8098a0d44f32ce259ff0487aec02cf</td><td>IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART B: CYBERNETICS, VOL. 36, NO. 4, AUGUST 2006 +<br/>863 +<br/>Bidirectional PCA With Assembled Matrix +<br/>Distance Metric for Image Recognition +</td></tr><tr><td>23ba9e462151a4bf9dfc3be5d8b12dbcfb7fe4c3</td><td>CS 229 Project, Fall 2014 +<br/>Determining Mood from Facial Expressions +<br/>Introduction +<br/>I +<br/>Facial expressions play an extremely important role in human communication. As +<br/>society continues to make greater use of human-machine interactions, it is important for +<br/>machines to be able to interpret facial expressions in order to improve their +<br/>authenticity. If machines can be trained to determine mood to a better extent than +<br/>humans can, especially for more subtle moods, then this could be useful in fields such as +<br/>counseling. This could also be useful for gauging reactions of large audiences in various +<br/>contexts, such as political talks. +<br/>The results of this project could also be applied to recognizing other features of facial +<br/>expressions, such as determining when people are purposefully suppressing emotions or +<br/>lying. The ability to recognize different facial expressions could also improve technology +<br/>that recognizes to whom specific faces belong. This could in turn be used to search a +<br/>large number of pictures for a specific photo, which is becoming increasingly difficult, as +<br/>storing photos digitally has been extremely common in the past decade. The possibilities +<br/>are endless. +<br/>II Data and Features +<br/>2.1 Data +<br/>Our data consists of 1166 frontal images of +<br/>people’s faces from three databases, with each +<br/>image labeled with one of eight emotions: +<br/>anger, contempt, disgust, fear, happiness, +<br/>neutral, sadness, and surprise. The TFEID [1], +<br/>CK+ [2], and JAFFE [3] databases primarily +<br/>consist of Taiwanese, Caucasian, and Japanese +<br/>subjects, respectively. The TFEID and JAFFE +<br/>images are both cropped with the faces +<br/>centered. Each image has a subject posing with +<br/>one of the emotions. The JAFFE database does +<br/>not have any images for contempt. +<br/>2.2 Features +<br/>On each face, there are many different facial landmarks. While some of these landmarks +<br/>(pupil position, nose tip, and face contour) are not as indicative of emotion, others +<br/>(eyebrow, mouth, and eye shape) are. To extract landmark data from images, we used +<br/>Happiness +<br/>Figure 1 +<br/>Anger +</td></tr><tr><td>238fc68b2e0ef9f5ec043d081451902573992a03</td><td>2656 +<br/>Enhanced Local Gradient Order Features and +<br/>Discriminant Analysis for Face Recognition +<br/>role in robust face recognition [5]. Many algorithms have +<br/>been proposed to deal with the effectiveness of feature design +<br/>and extraction [6], [7]; however, the performance of many +<br/>existing methods is still highly sensitive to variations of +<br/>imaging conditions, such as outdoor illumination, exaggerated +<br/>expression, and continuous occlusion. These complex varia- +<br/>tions are significantly affecting the recognition accuracy in +<br/>recent years [8]–[10]. +<br/>Appearance-based subspace learning is one of the sim- +<br/>plest approach for feature extraction, and many methods +<br/>are usually based on linear correlation of pixel intensities. +<br/>For example, Eigenface [11] uses eigen system of pixel +<br/>intensities to estimate the lower rank linear subspace of +<br/>a set of training face images by minimizing the (cid:2)2 dis- +<br/>tance metric. The solution enjoys optimality properties when +<br/>noise is independent +<br/>identically distributed Gaussian only. +<br/>Fisherface [12] will suffer more due to the estimation of +<br/>inverse within-class covariance matrix [13], +<br/>thus the per- +<br/>formance will degenerate rapidly in the cases of occlusion +<br/>and small sample size. Laplacianfaces [14] refer to another +<br/>appearance-based approach which learns a locality preserv- +<br/>ing subspace and seeks to capture the intrinsic geometry +<br/>and local structure of the data. Other methods such as those +<br/>in [5] and [15] also provide valuable approaches to supervised +<br/>or unsupervised dimension reduction tasks. +<br/>A fundamental problem of appearance-based methods for +<br/>face recognition, however, is that they are sensitive to imag- +<br/>ing conditions [10]. As for data corrupted by illumination +<br/>changes, occlusions, and inaccurate alignment, the estimated +<br/>subspace will be biased, thus much of the efforts concentrate +<br/>on removing/shrinking the noise components. In contrast, local +<br/>feature descriptors [15]–[19] have certain advantages as they +<br/>are more stable to local changes. In the view of image pro- +<br/>cessing and vision, the basic imaging system can be simply +<br/>formulated as +<br/>(x, y) = A(x, y) × L(x, y) +<br/>(1) +</td></tr><tr><td>23d55061f7baf2ffa1c847d356d8f76d78ebc8c1</td><td>Solmaz et al. IPSJ Transactions on Computer Vision and +<br/>Applications (2017) 9:22 +<br/>DOI 10.1186/s41074-017-0033-4 +<br/>IPSJ Transactions on Computer +<br/>Vision and Applications +<br/>RESEARCH PAPER +<br/>Open Access +<br/>Generic and attribute-specific deep +<br/>representations for maritime vessels +</td></tr><tr><td>23a8d02389805854cf41c9e5fa56c66ee4160ce3</td><td>Multimed Tools Appl +<br/>DOI 10.1007/s11042-013-1568-8 +<br/>Influence of low resolution of images on reliability +<br/>of face detection and recognition +<br/>© The Author(s) 2013. This article is published with open access at SpringerLink.com +</td></tr><tr><td>4fd29e5f4b7186e349ba34ea30738af7860cf21f</td><td></td></tr><tr><td>4f051022de100241e5a4ba8a7514db9167eabf6e</td><td>Face Parsing via a Fully-Convolutional Continuous +<br/>CRF Neural Network +</td></tr><tr><td>4f6adc53798d9da26369bea5a0d91ed5e1314df2</td><td>IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. , NO. , 2016 +<br/>Online Nonnegative Matrix Factorization with +<br/>General Divergences +</td></tr><tr><td>4fbef7ce1809d102215453c34bf22b5f9f9aab26</td><td></td></tr><tr><td>4fa0d73b8ba114578744c2ebaf610d2ca9694f45</td><td></td></tr><tr><td>4f591e243a8f38ee3152300bbf42899ac5aae0a5</td><td>SUBMITTED TO TPAMI +<br/>Understanding Higher-Order Shape +<br/>via 3D Shape Attributes +</td></tr><tr><td>4f9958946ad9fc71c2299847e9ff16741401c591</td><td>Facial Expression Recognition with Recurrent Neural Networks +<br/>Robotics and Embedded Systems Lab, Department of Computer Science +<br/>Image Understanding and Knowledge-Based Systems, Department of Computer Science +<br/>Technische Universit¨at M¨unchen, Germany +</td></tr><tr><td>4f0bf2508ae801aee082b37f684085adf0d06d23</td><td></td></tr><tr><td>4f4f920eb43399d8d05b42808e45b56bdd36a929</td><td>International Journal of Computer Applications (0975 – 8887) +<br/>Volume 123 – No.4, August 2015 +<br/>A Novel Method for 3D Image Segmentation with Fusion +<br/>of Two Images using Color K-means Algorithm +<br/>Neelam Kushwah +<br/>Dept. of CSE +<br/>ITM Universe +<br/>Gwalior +<br/>Priusha Narwariya +<br/>Dept. of CSE +<br/>ITM Universe +<br/>Gwalior +<br/>two +</td></tr><tr><td>8d71872d5877c575a52f71ad445c7e5124a4b174</td><td></td></tr><tr><td>8de06a584955f04f399c10f09f2eed77722f6b1c</td><td>Author manuscript, published in "International Conference on Computer Vision Theory and Applications (VISAPP 2013) (2013)" +</td></tr><tr><td>8d4f0517eae232913bf27f516101a75da3249d15</td><td>ARXIV SUBMISSION, MARCH 2018 +<br/>Event-based Dynamic Face Detection and +<br/>Tracking Based on Activity +</td></tr><tr><td>8de2dbe2b03be8a99628ffa000ac78f8b66a1028</td><td>´Ecole Nationale Sup´erieure dInformatique et de Math´ematiques Appliqu´ees de Grenoble +<br/>INP Grenoble – ENSIMAG +<br/>UFR Informatique et Math´ematiques Appliqu´ees de Grenoble +<br/>Rapport de stage de Master 2 et de projet de fin d’´etudes +<br/>Effectu´e au sein de l’´equipe LEAR, I.N.R.I.A., Grenoble +<br/>Action Recognition in Videos +<br/>3e ann´ee ENSIMAG – Option I.I.I. +<br/>M2R Informatique – sp´ecialit´e I.A. +<br/>04 f´evrier 2008 – 04 juillet 2008 +<br/>LEAR, +<br/>I.N.R.I.A., Grenoble +<br/>655 avenue de l’Europe +<br/>38 334 Montbonnot +<br/>France +<br/>Responsable de stage +<br/>Mme. Cordelia Schmid +<br/>Tuteur ´ecole +<br/>Jury +</td></tr><tr><td>8d42a24d570ad8f1e869a665da855628fcb1378f</td><td>CVPR +<br/>#987 +<br/>000 +<br/>001 +<br/>002 +<br/>003 +<br/>004 +<br/>005 +<br/>006 +<br/>007 +<br/>008 +<br/>009 +<br/>010 +<br/>011 +<br/>012 +<br/>013 +<br/>014 +<br/>015 +<br/>016 +<br/>017 +<br/>018 +<br/>019 +<br/>020 +<br/>021 +<br/>022 +<br/>023 +<br/>024 +<br/>025 +<br/>026 +<br/>027 +<br/>028 +<br/>029 +<br/>030 +<br/>031 +<br/>032 +<br/>033 +<br/>034 +<br/>035 +<br/>036 +<br/>037 +<br/>038 +<br/>039 +<br/>040 +<br/>041 +<br/>042 +<br/>043 +<br/>044 +<br/>045 +<br/>046 +<br/>047 +<br/>048 +<br/>049 +<br/>050 +<br/>051 +<br/>052 +<br/>053 +<br/>CVPR 2009 Submission #987. CONFIDENTIAL REVIEW COPY. DO NOT DISTRIBUTE. +<br/>An Empirical Study of Context in Object Detection +<br/>Anonymous CVPR submission +<br/>Paper ID 987 +</td></tr><tr><td>8d8461ed57b81e05cc46be8e83260cd68a2ebb4d</td><td>Age identification of Facial Images using Neural +<br/>Network +<br/>CSE Department,CSVTU +<br/>RIT, Raipur, Chhattisgarh , INDIA +</td></tr><tr><td>8dbe79830713925affc48d0afa04ed567c54724b</td><td></td></tr><tr><td>8d1adf0ac74e901a94f05eca2f684528129a630a</td><td>Facial Expression Recognition Using Facial +<br/>Movement Features +</td></tr><tr><td>8d712cef3a5a8a7b1619fb841a191bebc2a17f15</td><td></td></tr><tr><td>8dffbb6d75877d7d9b4dcde7665888b5675deee1</td><td>Emotion Recognition with Deep-Belief +<br/>Networks +<br/>Introduction +<br/>For our CS229 project, we studied the problem of +<br/>reliable computerized emotion recognition in images of +<br/>human +<br/>faces. First, we performed a preliminary +<br/>exploration using SVM classifiers, and then developed an +<br/>approach based on Deep Belief Nets. Deep Belief Nets, or +<br/>DBNs, are probabilistic generative models composed of +<br/>multiple layers of stochastic latent variables, where each +<br/>“building block” layer is a Restricted Boltzmann Machine +<br/>(RBM). DBNs have a greedy layer-wise unsupervised +<br/>learning algorithm as well as a discriminative fine-tuning +<br/>procedure for optimizing performance on classification +<br/>tasks. [1]. +<br/>We trained our classifier on three databases: the +<br/>Cohn-Kanade Extended Database (CK+) [2], the Japanese +<br/>Female Facial Expression Database (JAFFE) [3], and the +<br/>Yale Face Database (YALE) [4]. We tested several +<br/>different database configurations, image pre-processing +<br/>settings, and DBN parameters, and obtained test errors as +<br/>low as 20% on a limited subset of the emotion labels. +<br/>Finally, we created a real-time system which takes +<br/>images of a single subject using a computer webcam and +<br/>classifies the emotion shown by the subject. +<br/>Part 1: Exploration of SVM-based approaches +<br/>To set a baseline for comparison, we applied an +<br/>SVM classifier to the emotion images in the CK+ +<br/>database, using the LIBLINEAR library and its MATLAB +<br/>interface [5]. This database contains 593 image sequences +<br/>across 123 human subjects, beginning with a “neutral +<br/>“expression and showing the progression to one of seven +<br/>“peak” emotions. When given both a neutral and an +<br/>expressive face to compare, the SVM obtained accuracy +<br/>as high as 90%. This +<br/>the +<br/>implementation of the SVM classifier. For additional +<br/>details on this stage of the project, please see our +<br/>Milestone document. +<br/>Part 1.1 Choice of labels (emotion numbers vs. FACS +<br/>features) +<br/>The CK+ database offers two sets of emotion +<br/>features: “emotion numbers” and FACS features. Emotion +<br/>numbers are integer values representing the main emotion +<br/>shown in the “peak emotion” image. The emotions are +<br/>coded as follows: 1=anger, 2=contempt, 3=disgust, +<br/>4=fear, 5=happiness, 6=sadness, and 7=surprise. +<br/>The other labeling option is called FACS, or the +<br/>Facial Action Coding System. FACS decomposes every +<br/>summarizes +<br/>section +<br/>facial emotion into a set of Action Units (AUs), which +<br/>describe the specific muscle groups involved in forming +<br/>the emotion. We chose not to use FACS because accurate +<br/>labeling currently requires trained human experts [8], and +<br/>we are interesting in creating an automated system. +<br/> +<br/>Part 1.2 Features +<br/>Part 1.2.1 Norm of differences between neutral face +<br/>and full emotion +<br/>Each of the CK+ images has been hand-labeled with +<br/>68 standard Active Appearance Models (AAM) face +<br/>landmarks that describe the X and Y position of these +<br/>landmarks on the image (Figure 1). +<br/>Figure 1. AAM Facial Landmarks +<br/>We initially trained the SVM on the norm of the +<br/>vector differences in landmark positions between the +<br/>neutral and peak expressions. With this approach, the +<br/>training error was approximately 35% for hold out cross +<br/>validation (see Figure 2). +<br/>with +<br/>Figure 3. Accuracy of +<br/>SVM with separate X, Y +<br/>displacement features. +<br/>Figure 2. Accuracy of +<br/>SVM +<br/>norm- +<br/>displacement features. +<br/>Part 1.2.2 Separate X and Y differences between +<br/>neutral face and full emotion +<br/>Because the initial approach did not differentiate +<br/>between displacements of +<br/>in different +<br/>directions, we also provided the differences in the X and +<br/>Y components of each landmark separately. This doubled +<br/>the size of our feature vector, and resulting in a significant +<br/>(about 20%) improvement in accuracy (Figure 3). +<br/>Part 1.2.3 Feature Selection +<br/>landmarks +<br/>Finally, we visualized which features were the most +<br/>important for classifying each emotion; the results can be +<br/>seen in Figure 4. The figure shows the X and Y +</td></tr><tr><td>153f5ad54dd101f7f9c2ae17e96c69fe84aa9de4</td><td>Overview of algorithms for face detection and +<br/>tracking +<br/>Nenad Markuˇs +</td></tr><tr><td>15136c2f94fd29fc1cb6bedc8c1831b7002930a6</td><td>Deep Learning Architectures for Face +<br/>Recognition in Video Surveillance +</td></tr><tr><td>153e5cddb79ac31154737b3e025b4fb639b3c9e7</td><td>PREPRINT SUBMITTED TO IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS +<br/>Active Dictionary Learning in Sparse +<br/>Representation Based Classification +</td></tr><tr><td>15e0b9ba3389a7394c6a1d267b6e06f8758ab82b</td><td>Xu et al. IPSJ Transactions on Computer Vision and +<br/>Applications (2017) 9:24 +<br/>DOI 10.1186/s41074-017-0035-2 +<br/>IPSJ Transactions on Computer +<br/>Vision and Applications +<br/>TECHNICAL NOTE +<br/>Open Access +<br/>The OU-ISIR Gait Database comprising the +<br/>Large Population Dataset with Age and +<br/>performance evaluation of age estimation +</td></tr><tr><td>15aa6c457678e25f6bc0e818e5fc39e42dd8e533</td><td></td></tr><tr><td>15f3d47b48a7bcbe877f596cb2cfa76e798c6452</td><td>Automatic face analysis tools for interactive digital games +<br/>Anonymised for blind review +<br/>Anonymous +<br/>Anonymous +<br/>Anonymous +</td></tr><tr><td>15728d6fd5c9fc20b40364b733228caf63558c31</td><td></td></tr><tr><td>1513949773e3a47e11ab87d9a429864716aba42d</td><td></td></tr><tr><td>153c8715f491272b06dc93add038fae62846f498</td><td></td></tr><tr><td>122ee00cc25c0137cab2c510494cee98bd504e9f</td><td>The Application of +<br/>Active Appearance Models to +<br/>Comprehensive Face Analysis +<br/>Technical Report +<br/>TU M¨unchen +<br/>April 5, 2007 +</td></tr><tr><td>1287bfe73e381cc8042ac0cc27868ae086e1ce3b</td><td></td></tr><tr><td>12cb3bf6abf63d190f849880b1703ccc183692fe</td><td>Guess Who?: A game to crowdsource the labeling of affective facial +<br/>expressions is comparable to expert ratings. +<br/>Graduation research project, june 2012 +<br/>Supervised by: Dr. Joost Broekens +<br/><b></b></td></tr><tr><td>12cd96a419b1bd14cc40942b94d9c4dffe5094d2</td><td>29 +<br/>Proceedings of the 5th Workshop on Vision and Language, pages 29–38, +<br/>Berlin, Germany, August 12 2016. c(cid:13)2016 Association for Computational Linguistics +</td></tr><tr><td>12055b8f82d5411f9ad196b60698d76fbd07ac1e</td><td>1475 +<br/>Multiview Facial Landmark Localization in RGB-D +<br/>Images via Hierarchical Regression +<br/>With Binary Patterns +</td></tr><tr><td>120785f9b4952734818245cc305148676563a99b</td><td>Diagnostic automatique de l’état dépressif +<br/>S. Cholet +<br/>H. Paugam-Moisy +<br/>Laboratoire de Mathématiques Informatique et Applications (LAMIA - EA 4540) +<br/>Université des Antilles, Campus de Fouillole - Guadeloupe +<br/>Résumé +<br/>Les troubles psychosociaux sont un problème de santé pu- +<br/>blique majeur, pouvant avoir des conséquences graves sur +<br/>le court ou le long terme, tant sur le plan professionnel que +<br/>personnel ou familial. Le diagnostic de ces troubles doit +<br/>être établi par un professionnel. Toutefois, l’IA (l’Intelli- +<br/>gence Artificielle) peut apporter une contribution en four- +<br/>nissant au praticien une aide au diagnostic, et au patient +<br/>un suivi permanent rapide et peu coûteux. Nous proposons +<br/>une approche vers une méthode de diagnostic automatique +<br/>de l’état dépressif à partir d’observations du visage en +<br/>temps réel, au moyen d’une simple webcam. A partir de +<br/>vidéos du challenge AVEC’2014, nous avons entraîné un +<br/>classifieur neuronal à extraire des prototypes de visages +<br/>selon différentes valeurs du score de dépression de Beck +<br/>(BDI-II). +</td></tr><tr><td>12c713166c46ac87f452e0ae383d04fb44fe4eb2</td><td></td></tr><tr><td>12150d8b51a2158e574e006d4fbdd3f3d01edc93</td><td>Deep End2End Voxel2Voxel Prediction +<br/>Presented by: Ahmed Osman +<br/>Ahmed Osman +</td></tr><tr><td>8c13f2900264b5cf65591e65f11e3f4a35408b48</td><td>A GENERIC FACE REPRESENTATION APPROACH FOR +<br/>LOCAL APPEARANCE BASED FACE VERIFICATION +<br/>Interactive Systems Labs, Universität Karlsruhe (TH) +<br/>76131 Karlsruhe, Germany +<br/>web: http://isl.ira.uka.de/face_recognition/ +</td></tr><tr><td>8cb3f421b55c78e56c8a1c1d96f23335ebd4a5bf</td><td></td></tr><tr><td>8c955f3827a27e92b6858497284a9559d2d0623a</td><td>Buletinul Ştiinţific al Universităţii "Politehnica" din Timişoara +<br/>Seria ELECTRONICĂ şi TELECOMUNICAŢII +<br/>TRANSACTIONS on ELECTRONICS and COMMUNICATIONS +<br/>Tom 53(67), Fascicola 1-2, 2008 +<br/>Facial Expression Recognition under Noisy Environment +<br/>Using Gabor Filters +</td></tr><tr><td>8ce9b7b52d05701d5ef4a573095db66ce60a7e1c</td><td>Structured Sparse Subspace Clustering: A Joint +<br/>Affinity Learning and Subspace Clustering +<br/>Framework +</td></tr><tr><td>8c6c0783d90e4591a407a239bf6684960b72f34e</td><td>SESSION +<br/>KNOWLEDGE ENGINEERING AND +<br/>MANAGEMENT + KNOWLEDGE ACQUISITION +<br/>Chair(s) +<br/>TBA +<br/>Int'l Conf. Information and Knowledge Engineering | IKE'13 |1 </td></tr><tr><td>8509abbde2f4b42dc26a45cafddcccb2d370712f</td><td>Improving precision and recall of face recognition in SIPP with combination of +<br/>modified mean search and LSH +<br/>Xihua.Li +</td></tr><tr><td>855bfc17e90ec1b240efba9100fb760c068a8efa</td><td></td></tr><tr><td>858ddff549ae0a3094c747fb1f26aa72821374ec</td><td>Survey on RGB, 3D, Thermal, and Multimodal +<br/>Approaches for Facial Expression Recognition: +<br/>History, Trends, and Affect-related Applications +</td></tr><tr><td>858901405086056361f8f1839c2f3d65fc86a748</td><td>ON TENSOR TUCKER DECOMPOSITION: THE CASE FOR AN +<br/>ADJUSTABLE CORE SIZE +</td></tr><tr><td>85188c77f3b2de3a45f7d4f709b6ea79e36bd0d9</td><td>Author manuscript, published in "Workshop on Faces in 'Real-Life' Images: Detection, Alignment, and Recognition, Marseille : +<br/>France (2008)" +</td></tr><tr><td>8518b501425f2975ea6dcbf1e693d41e73d0b0af</td><td>Relative Hidden Markov Models for Evaluating Motion Skills +<br/>Computer Science and Engineering +<br/>Arizona State Univerisity, Tempe, AZ 85281 +</td></tr><tr><td>854dbb4a0048007a49df84e3f56124d387588d99</td><td>JOURNAL OF LATEX CLASS FILES, VOL. 13, NO. 9, SEPTEMBER 2014 +<br/>Spatial-Temporal Recurrent Neural Network for +<br/>Emotion Recognition +</td></tr><tr><td>1dbbec4ad8429788e16e9f3a79a80549a0d7ac7b</td><td></td></tr><tr><td>1d7ecdcb63b20efb68bcc6fd99b1c24aa6508de9</td><td>1860 +<br/>The Hidden Sides of Names—Face Modeling +<br/>with First Name Attributes +</td></tr><tr><td>1d846934503e2bd7b8ea63b2eafe00e29507f06a</td><td></td></tr><tr><td>1d0dd20b9220d5c2e697888e23a8d9163c7c814b</td><td>NEGREL ET AL.: BOOSTED METRIC LEARNING FOR FACE RETRIEVAL +<br/>Boosted Metric Learning for Efficient +<br/>Identity-Based Face Retrieval +<br/>Frederic Jurie +<br/>GREYC, CNRS UMR 6072, ENSICAEN +<br/>Université de Caen Basse-Normandie +<br/>France +</td></tr><tr><td>1d776bfe627f1a051099997114ba04678c45f0f5</td><td>Deployment of Customized Deep Learning based +<br/>Video Analytics On Surveillance Cameras +<br/>AitoeLabs (www.aitoelabs.com) +</td></tr><tr><td>1d3e01d5e2721dcfafe5a3b39c54ee1c980350bb</td><td></td></tr><tr><td>1de8f38c35f14a27831130060810cf9471a62b45</td><td>Int J Comput Vis +<br/>DOI 10.1007/s11263-017-0989-7 +<br/>A Branch-and-Bound Framework for Unsupervised Common +<br/>Event Discovery +<br/>Received: 3 June 2016 / Accepted: 12 January 2017 +<br/>© Springer Science+Business Media New York 2017 +</td></tr><tr><td>1da83903c8d476c64c14d6851c85060411830129</td><td>Iterated Support Vector Machines for Distance +<br/>Metric Learning +</td></tr><tr><td>1d6068631a379adbcff5860ca2311b790df3a70f</td><td></td></tr><tr><td>1d58d83ee4f57351b6f3624ac7e727c944c0eb8d</td><td>Enhanced Local Texture +<br/>Feature Sets for Face +<br/>Recognition under Difficult +<br/>Lighting Conditions +<br/>INRIA & Laboratoire Jean +<br/>Kuntzmann, +<br/>655 avenue de l'Europe, Montbonnot 38330, France +</td></tr><tr><td>71b376dbfa43a62d19ae614c87dd0b5f1312c966</td><td>The Temporal Connection Between Smiles and Blinks +</td></tr><tr><td>714d487571ca0d676bad75c8fa622d6f50df953b</td><td>eBear: An Expressive Bear-Like Robot +</td></tr><tr><td>710011644006c18291ad512456b7580095d628a2</td><td>Learning Residual Images for Face Attribute Manipulation +<br/>Fujitsu Research & Development Center, Beijing, China. +</td></tr><tr><td>76fd801981fd69ff1b18319c450cb80c4bc78959</td><td>Proceedings of the 11th International Conference on Computational Semantics, pages 76–81, +<br/>London, UK, April 15-17 2015. c(cid:13)2015 Association for Computational Linguistics +<br/>76 +</td></tr><tr><td>76dc11b2f141314343d1601635f721fdeef86fdb</td><td>Weighted Decoding ECOC for Facial +<br/>Action Unit Classification +</td></tr><tr><td>760a712f570f7a618d9385c0cee7e4d0d6a78ed2</td><td></td></tr><tr><td>76b9fe32d763e9abd75b427df413706c4170b95c</td><td></td></tr><tr><td>76d9f5623d3a478677d3f519c6e061813e58e833</td><td>FAST ALGORITHMS FOR THE GENERALIZED FOLEY-SAMMON +<br/>DISCRIMINANT ANALYSIS +</td></tr><tr><td>765b2cb322646c52e20417c3b44b81f89860ff71</td><td>PoseShop: Human Image Database +<br/>Construction and Personalized +<br/>Content Synthesis +</td></tr><tr><td>7644d90efef157e61fe4d773d8a3b0bad5feccec</td><td></td></tr><tr><td>760ba44792a383acd9ca8bef45765d11c55b48d4</td><td>~ +<br/>I . +<br/>INTRODUCTION AND BACKGROUND +<br/>The purpose of this article is to introduce the +<br/>reader to the basic principles of classification with +<br/>class-specific features. It is written both for readers +<br/>interested in only the basic concepts as well as those +<br/>interested in getting started in applying the method. +<br/>For in-depth coverage, the reader is referred to a more +<br/>detailed article [l]. +<br/>Class-Specific Classifier: +<br/>Avoiding the Curse of +<br/>Dimensionality +<br/>PAUL M. BAGGENSTOSS, Member. lEEE +<br/>US. Naval Undersea Warfare Center +<br/>This article describes a new probabilistic method called the +<br/>“class-specific method” (CSM). CSM has the potential to avoid +<br/>the “curse of dimensionality” which plagues most clmiiiers +<br/>which attempt to determine the decision boundaries in a +<br/>highdimensional featue space. In contrast, in CSM, it is possible +<br/>to build classifiers without a ” n o n feature space. Separate +<br/>Law-dimensional features seta may be de6ned for each class, while +<br/>the decision funetions are projected back to the common raw data +<br/>space. CSM eflectively extends the classical classification theory +<br/>to handle multiple feature spaw.. It is completely general, and +<br/>requires no s i m p l i n g assumption such as Gaussianity or that +<br/>data lies in linear subspaces. +<br/>Manuscript received September 26, 2W2; revised February 12, +<br/>2003. +<br/>This work was supported by the Office of Naval Research. +<br/>Author’s address: US. Naval Undersea Warfare Center, Newport +<br/>Classification is the process of assigning data +<br/>to one of a set of pre-determined class labels [2]. +<br/>Classification is a fundamental problem that has +<br/>to be solved if machines are to approximate the +<br/>human functions of recognizing sounds, images, or +<br/>other sensory inputs. This is why classification is so +<br/>important for automation in today’s commercial and +<br/>military arenas. +<br/>Many of us have first-hand knowledge of +<br/>successful automated recognition systems from +<br/>cameras that recognize faces in airports to computers +<br/>that can scan and read printed and handwritten text, +<br/>or systems that can recognize human speech. These +<br/>systems are becoming more and more reliable and +<br/>accurate. Given reasonably clean input data, the +<br/>performance is often quite good if not perfect. But +<br/>many of these systems fail in applications where +<br/>clean, uncorrupted data is not available or if the +<br/>problem is complicated by variability of conditions +<br/>or by proliferation of inputs from unknown sources. +<br/>In military environments, the targets to he recognized +<br/>are often uncooperative and hidden in clutter and +<br/>interference. In short, military uses of such systems +<br/>still fall far short of what a well-trained alert human +<br/>operator can achieve. +<br/>We are often perplexed by the wide gap of +<br/>as a car door slamming. From +<br/>performance between humans and automated systems. +<br/>Allow a human listener to hear two or three examples +<br/>of a sound-such +<br/>these few examples, the human can recognize +<br/>the sound again and not confuse it with similar +<br/>interfering sounds. But try the same experiment with +<br/>general-purpose classifiers using neural networks +<br/>and the story is quite different. Depending on the +<br/>problem, the automated system may require hundreds, +<br/>thousands, even millions of examples for training +<br/>before it becomes both robust and reliable. +<br/>Why? The answer lies in what is known as the +<br/>“curse of dimensionality.” General-purpose classifiers +<br/>need to extract a large number of measurements, +<br/>or features, from the data to account for all the +<br/>different possibilities of data types. The large +<br/>collection of features form a high-dimensional space +<br/>that the classifier has to sub-divide into decision +<br/>boundaries. It is well-known that the complexity of +<br/>a high-dimensional space increases exponentially +<br/>with the number of measurements [31-and +<br/>so does +<br/>the difficulty of finding the hest decision boundaries +<br/>from a fixed amount of training data. Unless a lot +<br/>EEE A&E SYSTEMS MAGAZINE VOL. 19, NO. 1 JANUARY 2004 PART 2: TUTORIALS-BAGGENSTOSS +<br/>37 +</td></tr><tr><td>766728bac030b169fcbc2fbafe24c6e22a58ef3c</td><td>A survey of deep facial landmark detection +<br/>Yongzhe Yan1,2 +<br/>Thierry Chateau1 +<br/>1 Université Clermont Auvergne, France +<br/>2 Wisimage, France +<br/>3 Université de Lyon, CNRS, INSA Lyon, LIRIS, UMR5205, Lyon, France +<br/>Résumé +<br/>La détection de landmarks joue un rôle crucial dans de +<br/>nombreuses applications d’analyse du visage comme la +<br/>reconnaissance de l’identité, des expressions, l’animation +<br/>d’avatar, la reconstruction 3D du visage, ainsi que pour +<br/>les applications de réalité augmentée comme la pose de +<br/>masque ou de maquillage virtuel. L’avènement de l’ap- +<br/>prentissage profond a permis des progrès très importants +<br/>dans ce domaine, y compris sur les corpus non contraints +<br/>(in-the-wild). Nous présentons ici un état de l’art cen- +<br/>tré sur la détection 2D dans une image fixe, et les mé- +<br/>thodes spécifiques pour la vidéo. Nous présentons ensuite +<br/>les corpus existants pour ces trois tâches, ainsi que les mé- +<br/>triques d’évaluations associées. Nous exposons finalement +<br/>quelques résultats, ainsi que quelques pistes de recherche. +<br/>Mots Clef +<br/>Détection de landmark facial, Alignement de visage, Deep +<br/>learning +</td></tr><tr><td>7697295ee6fc817296bed816ac5cae97644c2d5b</td><td>Detecting and Recognizing Human-Object Interactions +<br/>Facebook AI Research (FAIR) +</td></tr><tr><td>1c80bc91c74d4984e6422e7b0856cf3cf28df1fb</td><td>Noname manuscript No. +<br/>(will be inserted by the editor) +<br/>Hierarchical Adaptive Structural SVM for Domain Adaptation +<br/>Received: date / Accepted: date +</td></tr><tr><td>1ce4587e27e2cf8ba5947d3be7a37b4d1317fbee</td><td>Deep fusion of visual signatures +<br/>for client-server facial analysis +<br/>Normandie Univ, UNICAEN, +<br/>ENSICAEN, CNRS, GREYC +<br/>Computer Sc. & Engg. +<br/>IIT Kanpur, India +<br/>Frederic Jurie +<br/>Normandie Univ, UNICAEN, +<br/>ENSICAEN, CNRS, GREYC +<br/>Facial analysis is a key technology for enabling human- +<br/>machine interaction. +<br/>In this context, we present a client- +<br/>server framework, where a client transmits the signature of +<br/>a face to be analyzed to the server, and, in return, the server +<br/>sends back various information describing the face e.g. is the +<br/>person male or female, is she/he bald, does he have a mus- +<br/>tache, etc. We assume that a client can compute one (or a +<br/>combination) of visual features; from very simple and effi- +<br/>cient features, like Local Binary Patterns, to more complex +<br/>and computationally heavy, like Fisher Vectors and CNN +<br/>based, depending on the computing resources available. The +<br/>challenge addressed in this paper is to design a common uni- +<br/>versal representation such that a single merged signature is +<br/>transmitted to the server, whatever be the type and num- +<br/>ber of features computed by the client, ensuring nonetheless +<br/>an optimal performance. Our solution is based on learn- +<br/>ing of a common optimal subspace for aligning the different +<br/>face features and merging them into a universal signature. +<br/>We have validated the proposed method on the challenging +<br/>CelebA dataset, on which our method outperforms existing +<br/>state-of-art methods when rich representation is available at +<br/>test time, while giving competitive performance when only +<br/>simple signatures (like LBP) are available at test time due +<br/>to resource constraints on the client. +<br/>1. +<br/>INTRODUCTION +<br/>We propose a novel method in a heterogeneous server- +<br/>client framework for the challenging and important task of +<br/>analyzing images of faces. Facial analysis is a key ingredient +<br/>for assistive computer vision and human-machine interaction +<br/>methods, and systems and incorporating high-performing +<br/>methods in daily life devices is a challenging task. The ob- +<br/>jective of the present paper is to develop state-of-the-art +<br/>technologies for recognizing facial expressions and facial at- +<br/>tributes on mobile and low cost devices. Depending on their +<br/>computing resources, the clients (i.e. the devices on which +<br/>the face image is taken) are capable of computing different +<br/>types of face signatures, from the simplest ones (e.g. LPB) +<br/>to the most complex ones (e.g. very deep CNN features), and +<br/>should be able to eventually combine them into a single rich +<br/>signature. Moreover, it is convenient if the face analyzer, +<br/>which might require significant computing resources, is im- +<br/>plemented on a server receiving face signatures and comput- +<br/>ing facial expressions and attributes from these signatures. +<br/>Keeping the computation of the signatures on the client is +<br/>safer in terms of privacy, as the original images are not trans- +<br/>mitted, and keeping the analysis part on the server is also +<br/>beneficial for easy model upgrades in the future. To limit +<br/>the transmission costs, the signatures have to be made as +<br/>compact as possible. +<br/>In summary, the technology needed +<br/>for this scenario has to be able to merge the different avail- +<br/>able features – the number of features available at test time +<br/>is not known in advance but is dependent on the computing +<br/>resources available on the client – producing a unique rich +<br/>and compact signature of the face, which can be transmitted +<br/>and analyzed by a server. Ideally, we would like the univer- +<br/>sal signature to have the following properties: when all the +<br/>features are available, we would like the performance of the +<br/>signature to be better than the one of a system specifically +<br/>optimized for any single type of feature. +<br/>In addition, we +<br/>would like to have reasonable performance when only one +<br/>type of feature is available at test time. +<br/>For developing such a system, we propose a hybrid deep +<br/>neural network and give a method to carefully fine-tune the +<br/>network parameters while learning with all or a subset of +<br/>features available. Thus, the proposed network can process a +<br/>number of wide ranges of feature types such as hand-crafted +<br/>LBP and FV, or even CNN features which are learned end- +<br/>to-end. +<br/>While CNNs have been quite successful in computer vi- +<br/>sion [1], representing images with CNN features is relatively +<br/>time consuming, much more than some simple hand-crafted +<br/>features such as LBP. Thus, the use of CNN in real-time ap- +<br/>plications is still not feasible. In addition, the use of robust +<br/>hand-crafted features such as FV in hybrid architectures can +<br/>give performance comparable to Deep CNN features [2]. The +<br/>main advantage of learning hybrid architectures is to avoid +<br/>having large numbers of convolutional and pooling layers. +<br/>Again from [2], we can also observe that hybrid architec- +<br/>tures improve the performance of hand-crafted features e.g. +<br/>FVs. Therefore, hybrid architectures are useful for the cases +<br/>where only hand-crafted features, and not the original im- +<br/>ages, are available during training and testing time. This +<br/>scenario is useful when it is not possible to share training +<br/>images due to copyright or privacy issues. +<br/>Hybrid networks are particularly adapted to our client- +</td></tr><tr><td>1c3073b57000f9b6dbf1c5681c52d17c55d60fd7</td><td>THÈSEprésentéepourl’obtentiondutitredeDOCTEURDEL’ÉCOLENATIONALEDESPONTSETCHAUSSÉESSpécialité:InformatiqueparCharlotteGHYSAnalyse,Reconstruction3D,&AnimationduVisageAnalysis,3DReconstruction,&AnimationofFacesSoutenancele19mai2010devantlejurycomposéde:Rapporteurs:MajaPANTICDimitrisSAMARASExaminateurs:MichelBARLAUDRenaudKERIVENDirectiondethèse:NikosPARAGIOSBénédicteBASCLE </td></tr><tr><td>1c93b48abdd3ef1021599095a1a5ab5e0e020dd5</td><td>JOURNAL OF LATEX CLASS FILES, VOL. *, NO. *, JANUARY 2009 +<br/>A Compositional and Dynamic Model for Face Aging +</td></tr><tr><td>1c6be6874e150898d9db984dd546e9e85c85724e</td><td></td></tr><tr><td>1c65f3b3c70e1ea89114f955624d7adab620a013</td><td></td></tr><tr><td>1c6e22516ceb5c97c3caf07a9bd5df357988ceda</td><td></td></tr><tr><td>82bef8481207de9970c4dc8b1d0e17dced706352</td><td></td></tr><tr><td>825f56ff489cdd3bcc41e76426d0070754eab1a8</td><td>Making Convolutional Networks Recurrent for Visual Sequence Learning +<br/>NVIDIA +</td></tr><tr><td>82d2af2ffa106160a183371946e466021876870d</td><td>A Novel Space-Time Representation on the Positive Semidefinite Cone +<br/>for Facial Expression Recognition +<br/>1IMT Lille Douai, Univ. Lille, CNRS, UMR 9189 – CRIStAL – +<br/>Centre de Recherche en Informatique Signal et Automatique de Lille, F-59000 Lille, France +<br/>2Univ. Lille, CNRS, UMR 8524, Laboratoire Paul Painlev´e, F-59000 Lille, France. +</td></tr><tr><td>8210fd10ef1de44265632589f8fc28bc439a57e6</td><td>Single Sample Face Recognition via Learning Deep +<br/>Supervised Auto-Encoders +<br/>Shenghua Gao, Yuting Zhang, Kui Jia, Jiwen Lu, Yingying Zhang +</td></tr><tr><td>82a4a35b2bae3e5c51f4d24ea5908c52973bd5be</td><td>Real-time emotion recognition for gaming using +<br/>deep convolutional network features +<br/>S´ebastien Ouellet +</td></tr><tr><td>82f4e8f053d20be64d9318529af9fadd2e3547ef</td><td>Technical Report: +<br/>Multibiometric Cryptosystems +</td></tr><tr><td>82d781b7b6b7c8c992e0cb13f7ec3989c8eafb3d</td><td>141 +<br/>REFERENCES +<br/>1. +<br/>2. +<br/>3. +<br/>4. +<br/>5. +<br/>6. +<br/>7. +<br/>8. +<br/>9. +<br/>Adler A., Youmaran R. and Loyka S., “Towards a Measure of +<br/>Biometric Information”, Canadian Conference on Electrical and +<br/>Computer Engineering, pp. 210-213, 2006. +<br/>Military Academy, West Point, New York, pp. 452-458, 2005. +<br/>Security and Trust, St. Andrews, New Brunswick, Canada, pp. 1-8, +<br/>2005. +<br/>Structural Model for Biometric Sketch Recognition”, Proceedings of +<br/>DAGM, Magdeburg, Germany, Vol. 2781, pp. 187-195, 2003. +<br/>of Security”, The First UAE International Conference on Biological +<br/>and Medical Physics, pp. 1-4, 2005. +<br/>Avraam Kasapis., “MLPs and Pose, Expression Classification”, +<br/>Proceedings of UNiS Report, pp. 1-87, 2003. +<br/>Detection for Storage Area Networks (SANs)”, Proceedings of 22nd +<br/>IEEE / 13th NASA Goddard Conference on Mass Storage Systems and +<br/>Technologies, pp. 118-127, 2005. +<br/>Black M.J. and Yacoob Y., “Recognizing Facial Expressions in Image +<br/>Sequences using Local Parameterized Models of Image Motion”, Int. +<br/>Journal Computer Vision, Vol. 25, No. 1, pp. 23-48, 1997. +<br/>10. +<br/>Recognition using a State-Based Model of Spatially-Localized Facial +</td></tr><tr><td>82417d8ec8ac6406f2d55774a35af2a1b3f4b66e</td><td>Some faces are more equal than others: +<br/>Hierarchical organization for accurate and +<br/>efficient large-scale identity-based face retrieval +<br/>GREYC, CNRS UMR 6072, Universit´e de Caen Basse-Normandie, France1 +<br/>Technicolor, Rennes, France2 +</td></tr><tr><td>826c66bd182b54fea3617192a242de1e4f16d020</td><td>978-1-5090-4117-6/17/$31.00 ©2017 IEEE +<br/>1602 +<br/>ICASSP 2017 +</td></tr><tr><td>4972aadcce369a8c0029e6dc2f288dfd0241e144</td><td>Multi-target Unsupervised Domain Adaptation +<br/>without Exactly Shared Categories +</td></tr><tr><td>49dd4b359f8014e85ed7c106e7848049f852a304</td><td></td></tr><tr><td>49e85869fa2cbb31e2fd761951d0cdfa741d95f3</td><td>253 +<br/>Adaptive Manifold Learning +</td></tr><tr><td>49659fb64b1d47fdd569e41a8a6da6aa76612903</td><td></td></tr><tr><td>49a7949fabcdf01bbae1c2eb38946ee99f491857</td><td>A CONCATENATING FRAMEWORK OF SHORTCUT +<br/>CONVOLUTIONAL NEURAL NETWORKS +</td></tr><tr><td>49e1aa3ecda55465641b2c2acc6583b32f3f1fc6</td><td>International Journal of Emerging Technology and Advanced Engineering +<br/>Website: www.ijetae.com (ISSN 2250-2459, Volume 2, Issue 5, May 2012) +<br/>Support Vector Machine for age classification +<br/>1Assistant Professor, CSE, RSR RCET, Kohka Bhilai +<br/>2,3 Sr. Assistant Professor, CSE, SSCET, Junwani Bhilai +</td></tr><tr><td>49df381ea2a1e7f4059346311f1f9f45dd997164</td><td>2018 +<br/>On the Use of Client-Specific Information for Face +<br/>Presentation Attack Detection Based on Anomaly +<br/>Detection +</td></tr><tr><td>40205181ed1406a6f101c5e38c5b4b9b583d06bc</td><td>Using Context to Recognize People in Consumer Images +</td></tr><tr><td>40dab43abef32deaf875c2652133ea1e2c089223</td><td>Noname manuscript No. +<br/>(will be inserted by the editor) +<br/>Facial Communicative Signals +<br/>Valence Recognition in Task-Oriented Human-Robot Interaction +<br/>Received: date / Accepted: date +</td></tr><tr><td>405b43f4a52f70336ac1db36d5fa654600e9e643</td><td>What can we learn about CNNs from a large scale controlled object dataset? +<br/>UWM +<br/>AUT +<br/>USC +</td></tr><tr><td>40b86ce698be51e36884edcc8937998979cd02ec</td><td>Yüz ve İsim İlişkisi kullanarak Haberlerdeki Kişilerin Bulunması +<br/>Finding Faces in News Photos Using Both Face and Name Information +<br/>Derya Ozkan, Pınar Duygulu +<br/>Bilgisayar Mühendisliği Bölümü, Bilkent Üniversitesi, 06800, Ankara +<br/>Özetçe +<br/>Bu çalışmada, haber fotoğraflarından oluşan geniş veri +<br/>kümelerinde kişilerin sorgulanmasını sağlayan bir yöntem +<br/>sunulmuştur. Yöntem isim ve yüzlerin ilişkilendirilmesine +<br/>dayanmaktadır. Haber başlığında kişinin ismi geçiyor ise +<br/>fotoğrafta da o kişinin yüzünün bulunacağı varsayımıyla, ilk +<br/>olarak sorgulanan isim ile ilişkilendirilmiş, fotoğraflardaki +<br/>tüm yüzler seçilir. Bu yüzler arasında sorgu kişisine ait farklı +<br/>koşul, poz ve zamanlarda çekilmiş pek çok resmin yanında, +<br/>haberde ismi geçen başka kişilere ait yüzler ya da kullanılan +<br/>yüz bulma yönteminin hatasından kaynaklanan yüz olmayan +<br/>resimler de bulunabilir. Yine de, çoğu zaman, sorgu kişisine +<br/>ait resimler daha çok olup, bu resimler birbirine diğerlerine +<br/>olduğundan daha çok benzeyeceklerdir. Bu nedenle, yüzler +<br/>arasındaki benzerlikler çizgesel olarak betimlendiğinde , +<br/>birbirine en çok benzeyen yüzler bu çizgede en yoğun bileşen +<br/>olacaktır. Bu çalışmada, sorgu ismiyle ilişkilendirilmiş, +<br/>yüzler arasında birbirine en çok benzeyen alt kümeyi bulan, +<br/>çizgeye dayalı bir yöntem sunulmaktadır. +</td></tr><tr><td>402f6db00251a15d1d92507887b17e1c50feebca</td><td>3D Facial Action Units Recognition for Emotional +<br/>Expression +<br/>1Department of Information Technology and Communication, Politeknik Kuching, Sarawak, Malaysia +<br/>2Faculty of Computer Science and Information Technology, Universiti Malaysia Sarawak, Kota Samarahan, Sarawak, Malaysia +<br/>The muscular activities caused the activation of certain AUs for every facial expression at the certain duration of time +<br/>throughout the facial expression. This paper presents the methods to recognise facial Action Unit (AU) using facial distance +<br/>of the facial features which activates the muscles. The seven facial action units involved are AU1, AU4, AU6, AU12, AU15, +<br/>AU17 and AU25 that characterises happy and sad expression. The recognition is performed on each AU according to rules +<br/>defined based on the distance of each facial points. The facial distances chosen are extracted from twelve facial features. +<br/>Then the facial distances are trained using Support Vector Machine (SVM) and Neural Network (NN). Classification result +<br/>using SVM is presented with several different SVM kernels while result using NN is presented for each training, validation +<br/>and testing phase. +<br/>Keywords: Facial action units recognition, 3D AU recognition, facial expression +<br/> +</td></tr><tr><td>40fb4e8932fb6a8fef0dddfdda57a3e142c3e823</td><td>A Mixed Generative-Discriminative Framework for Pedestrian Classification +<br/>Dariu M. Gavrila2,3 +<br/>1 Image & Pattern Analysis Group, Dept. of Math. and Comp. Sc., Univ. of Heidelberg, Germany +<br/>2 Environment Perception, Group Research, Daimler AG, Ulm, Germany +<br/>3 Intelligent Systems Lab, Faculty of Science, Univ. of Amsterdam, The Netherlands +</td></tr><tr><td>40cd062438c280c76110e7a3a0b2cf5ef675052c</td><td></td></tr><tr><td>40a1935753cf91f29ffe25f6c9dde2dc49bf2a3a</td><td>80 +</td></tr><tr><td>40a34d4eea5e32dfbcef420ffe2ce7c1ee0f23cd</td><td>Bridging Heterogeneous Domains With Parallel Transport For Vision and +<br/>Multimedia Applications +<br/>Dept. of Video and Multimedia Technologies Research +<br/>AT&T Labs-Research +<br/>San Francisco, CA 94108 +</td></tr><tr><td>40389b941a6901c190fb74e95dc170166fd7639d</td><td>Automatic Facial Expression Recognition +<br/>Emotient +<br/>http://emotient.com +<br/>February 12, 2014 +<br/>Imago animi vultus est, indices oculi. (Cicero) +<br/>Introduction +<br/>The face is innervated by two different brain systems that compete for control of its muscles: +<br/>a cortical brain system related to voluntary and controllable behavior, and a sub-cortical +<br/>system responsible for involuntary expressions. The interplay between these two systems +<br/>generates a wealth of information that humans constantly use to read the emotions, inten- +<br/>tions, and interests [25] of others. +<br/>Given the critical role that facial expressions play in our daily life, technologies that can +<br/>interpret and respond to facial expressions automatically are likely to find a wide range of +<br/>applications. For example, in pharmacology, the effect of new anti-depression drugs could +<br/>be assessed more accurately based on daily records of the patients’ facial expressions than +<br/>asking the patients to fill out a questionnaire, as it is currently done [7]. Facial expression +<br/>recognition may enable a new generation of teaching systems to adapt to the expression +<br/>of their students in the way good teachers do [61]. Expression recognition could be used +<br/>to assess the fatigue of drivers and air-pilots [58, 59]. Daily-life robots with automatic +<br/>expression recognition will be able to assess the states and intentions of humans and respond +<br/>accordingly [41]. Smart phones with expression analysis may help people to prepare for +<br/>important meetings and job interviews. +<br/>Thanks to the introduction of machine learning methods, recent years have seen great +<br/>progress in the field of automatic facial expression recognition. Commercial real-time ex- +<br/>pression recognition systems are starting to be used in consumer applications, e.g., smile +<br/>detectors embedded in digital cameras [62]. Nonetheless, considerable progress has yet to be +<br/>made: Methods for face detection and tracking (the first step of automated face analysis) +<br/>work well for frontal views of adult Caucasian and Asian faces [50], but their performance +</td></tr><tr><td>40273657e6919455373455bd9a5355bb46a7d614</td><td>Anonymizing k-Facial Attributes via Adversarial Perturbations +<br/>1 IIIT Delhi, New Delhi, India +<br/>2 Ministry of Electronics and Information Technology, New Delhi, India +</td></tr><tr><td>40b10e330a5511a6a45f42c8b86da222504c717f</td><td>Implementing the Viola-Jones +<br/>Face Detection Algorithm +<br/>Kongens Lyngby 2008 +<br/>IMM-M.Sc.-2008-93 +</td></tr><tr><td>40ca925befa1f7e039f0cd40d57dbef6007b4416</td><td>Sampling Matters in Deep Embedding Learning +<br/>UT Austin +<br/>A9/Amazon +<br/>Amazon +<br/>Philipp Kr¨ahenb¨uhl +<br/>UT Austin +</td></tr><tr><td>4042bbb4e74e0934f4afbedbe92dd3e37336b2f4</td><td></td></tr><tr><td>40f127fa4459a69a9a21884ee93d286e99b54c5f</td><td>Optimizing Apparent Display Resolution +<br/>Enhancement for Arbitrary Videos +</td></tr><tr><td>401e6b9ada571603b67377b336786801f5b54eee</td><td>Active Image Clustering: Seeking Constraints from +<br/>Humans to Complement Algorithms +<br/>November 22, 2011 +</td></tr><tr><td>2e20ed644e7d6e04dd7ab70084f1bf28f93f75e9</td><td></td></tr><tr><td>2e8e6b835e5a8f55f3b0bdd7a1ff765a0b7e1b87</td><td>International Journal of Computer Vision manuscript No. +<br/>(will be inserted by the editor) +<br/>Pointly-Supervised Action Localization +<br/>Received: date / Accepted: date +</td></tr><tr><td>2eb37a3f362cffdcf5882a94a20a1212dfed25d9</td><td>4 +<br/>Local Feature Based Face Recognition +<br/>R.I.T., Rajaramnagar and S.G.G.S. COE &T, Nanded +<br/>India +<br/>1. Introduction +<br/>A reliable automatic face recognition (AFR) system is a need of time because in today's +<br/>networked world, maintaining the security of private information or physical property is +<br/>becoming increasingly important and difficult as well. Most of the time criminals have been +<br/>taking the advantage of fundamental flaws in the conventional access control systems i.e. +<br/>the systems operating on credit card, ATM etc. do not grant access by "who we are", but by +<br/>"what we have”. The biometric based access control systems have a potential to overcome +<br/>most of the deficiencies of conventional access control systems and has been gaining the +<br/>importance in recent years. These systems can be designed with biometric traits such as +<br/>fingerprint, face, iris, signature, hand geometry etc. But comparison of different biometric +<br/>traits shows that face is very attractive biometric because of its non-intrusiveness and social +<br/>acceptability. It provides automated methods of verifying or recognizing the identity of a +<br/>living person based on its facial characteristics. +<br/>In last decade, major advances occurred in face recognition, with many systems capable of +<br/>achieving recognition rates greater than 90%. However real-world scenarios remain a +<br/>challenge, because face acquisition process can undergo to a wide range of variations. Hence +<br/>the AFR can be thought as a very complex object recognition problem, where the object to be +<br/>recognized is the face. This problem becomes even more difficult because the search is done +<br/>among objects belonging to the same class and very few images of each class are available to +<br/>train the system. Moreover different problems arise when images are acquired under +<br/>uncontrolled conditions such as illumination variations, pose changes, occlusion, person +<br/>appearance at different ages, expression changes and face deformations. The numbers of +<br/>approaches has been proposed by various researchers to deal with these problems but still +<br/>reported results cannot suffice the need of the reliable AFR system in presence of all facial +<br/>image variations. A recent survey paper (Abate et al., 2007) states that the sensibility of the +<br/>AFR systems to illumination and pose variations are the main problems researchers have +<br/>been facing up till. +<br/>2. Face recognition methods +<br/>The existing face recognition methods can be divided into two categories: holistic matching +<br/>methods and local matching methods.The holistic matching methods use complete face +<br/>region as a input to face recognition system and constructs a lower dimensional subspace +<br/>using principal component analysis (PCA) (Turk & Pentland, 1991), linear discriminant +<br/>www.intechopen.com +</td></tr><tr><td>2e5cfa97f3ecc10ae8f54c1862433285281e6a7c</td><td></td></tr><tr><td>2e0e056ed5927a4dc6e5c633715beb762628aeb0</td><td></td></tr><tr><td>2e68190ebda2db8fb690e378fa213319ca915cf8</td><td>Generating Videos with Scene Dynamics +<br/>MIT +<br/>UMBC +<br/>MIT +</td></tr><tr><td>2e0d56794379c436b2d1be63e71a215dd67eb2ca</td><td>Improving precision and recall of face recognition in SIPP with combination of +<br/>modified mean search and LSH +<br/>Xihua.Li +</td></tr><tr><td>2ee8900bbde5d3c81b7ed4725710ed46cc7e91cd</td><td></td></tr><tr><td>2ef51b57c4a3743ac33e47e0dc6a40b0afcdd522</td><td>Leveraging Billions of Faces to Overcome +<br/>Performance Barriers in Unconstrained Face +<br/>Recognition +<br/>face.com +</td></tr><tr><td>2e19371a2d797ab9929b99c80d80f01a1fbf9479</td><td></td></tr><tr><td>2ebc35d196cd975e1ccbc8e98694f20d7f52faf3</td><td>This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. +<br/>IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE +<br/>Towards Wide-angle Micro Vision Sensors +</td></tr><tr><td>2e3d081c8f0e10f138314c4d2c11064a981c1327</td><td></td></tr><tr><td>2e86402b354516d0a8392f75430156d629ca6281</td><td></td></tr><tr><td>2e0f5e72ad893b049f971bc99b67ebf254e194f7</td><td>Apparel Classification with Style +<br/>1ETH Z¨urich, Switzerland 2Microsoft, Austria 3Kooaba AG, Switzerland +<br/>4KU Leuven, Belgium +</td></tr><tr><td>2ec7d6a04c8c72cc194d7eab7456f73dfa501c8c</td><td>International Journal of Scientific Research and Management Studies (IJSRMS) +<br/>ISSN: 2349-3771 +<br/> +<br/>Volume 3 Issue 4, pg: 164-169 +<br/>A REVIEW ON TEXTURE BASED EMOTION RECOGNITION +<br/>FROM FACIAL EXPRESSION +<br/>1U.G. Scholars, 2Assistant Professor, +<br/>Dept. of E & C Engg., MIT Moradabad, Ram Ganga Vihar, Phase II, Moradabad, India. +</td></tr><tr><td>2e1b1969ded4d63b69a5ec854350c0f74dc4de36</td><td></td></tr><tr><td>2b0ff4b82bac85c4f980c40b3dc4fde05d3cc23f</td><td>An Effective Approach for Facial Expression Recognition with Local Binary +<br/>Pattern and Support Vector Machine +</td></tr><tr><td>2b3ceb40dced78a824cf67054959e250aeaa573b</td><td></td></tr><tr><td>2b1327a51412646fcf96aa16329f6f74b42aba89</td><td>Under review as a conference paper at ICLR 2016 +<br/>IMPROVING PERFORMANCE OF RECURRENT NEURAL +<br/>NETWORK WITH RELU NONLINEARITY +<br/>Qualcomm Research +<br/>San Diego, CA 92121, USA +</td></tr><tr><td>2b5cb5466eecb131f06a8100dcaf0c7a0e30d391</td><td>A Comparative Study of Active Appearance Model +<br/>Annotation Schemes for the Face +<br/>Face Aging Group +<br/>UNCW, USA +<br/>Face Aging Group +<br/>UNCW, USA +<br/>Face Aging Group +<br/>UNCW, USA +</td></tr><tr><td>2b632f090c09435d089ff76220fd31fd314838ae</td><td>Early Adaptation of Deep Priors in Age Prediction from Face Images +<br/>Computer Vision Lab +<br/>D-ITET, ETH Zurich +<br/>Computer Vision Lab +<br/>D-ITET, ETH Zurich +<br/>CVL, D-ITET, ETH Zurich +<br/>Merantix GmbH +</td></tr><tr><td>2b8dfbd7cae8f412c6c943ab48c795514d53c4a7</td><td>529 +<br/>2014 IEEE International Conference on Acoustic, Speech and Signal Processing (ICASSP) +<br/>978-1-4799-2893-4/14/$31.00 ©2014 IEEE +<br/>RECOGNITION +<br/>1. INTRODUCTION +<br/>(d1,d2)∈[0;d]2 +<br/>d1+d2≤d +</td></tr><tr><td>2baec98c19804bf19b480a9a0aa814078e28bb3d</td><td></td></tr><tr><td>470dbd3238b857f349ebf0efab0d2d6e9779073a</td><td>Unsupervised Simultaneous Orthogonal Basis Clustering Feature Selection +<br/>School of Electrical Engineering, KAIST, South Korea +<br/>In this paper, we propose a novel unsupervised feature selection method: Si- +<br/>multaneous Orthogonal basis Clustering Feature Selection (SOCFS). To per- +<br/>form feature selection on unlabeled data effectively, a regularized regression- +<br/>based formulation with a new type of target matrix is designed. The target +<br/>matrix captures latent cluster centers of the projected data points by per- +<br/>forming the orthogonal basis clustering, and then guides the projection ma- +<br/>trix to select discriminative features. Unlike the recent unsupervised feature +<br/>selection methods, SOCFS does not explicitly use the pre-computed local +<br/>structure information for data points represented as additional terms of their +<br/>objective functions, but directly computes latent cluster information by the +<br/>target matrix conducting orthogonal basis clustering in a single unified term +<br/>of the proposed objective function. +<br/>Since the target matrix is put in a single unified term for regression of +<br/>the proposed objective function, feature selection and clustering are simul- +<br/>taneously performed. In this way, the projection matrix for feature selection +<br/>is more properly computed by the estimated latent cluster centers of the +<br/>projected data points. To the best of our knowledge, this is the first valid +<br/>formulation to consider feature selection and clustering together in a sin- +<br/>gle unified term of the objective function. The proposed objective function +<br/>has fewer parameters to tune and does not require complicated optimization +<br/>tools so just a simple optimization algorithm is sufficient. Substantial ex- +<br/>periments are performed on several publicly available real world datasets, +<br/>which shows that SOCFS outperforms various unsupervised feature selec- +<br/>tion methods and that latent cluster information by the target matrix is ef- +<br/>fective for regularized regression-based feature selection. +<br/>Problem Formulation: Given training data, let X = [x1, . . . ,xn] ∈ Rd×n +<br/>denote the data matrix with n instances where dimension is d and T = +<br/>[t1, . . . ,tn] ∈ Rm×n denote the corresponding target matrix where dimension +<br/>is m. We start from the regularized regression-based formulation to select +<br/>maximum r features is minW (cid:107)WT X− T(cid:107)2 +<br/>s.t. (cid:107)W(cid:107)2,0 ≤ r. To exploit +<br/>such formulation on unlabeled data more effectively, it is crucial for the tar- +<br/>get matrix T to have discriminative destinations for projected clusters. To +<br/>this end, a new type of target matrix T is proposed to conduct clustering di- +<br/>rectly on the projected data points WT X. We allow extra degrees of freedom +<br/>to T by decomposing it into two other matrices B ∈ Rm×c and E ∈ Rn×c as +<br/>T = BET with additional constraints as +<br/>(1) +<br/>F + λ(cid:107)W(cid:107)2,1 +<br/>(cid:107)WT X− BET(cid:107)2 +<br/>s.t. BT B = I, ET E = I, E ≥ 0, +<br/>min +<br/>W,B,E +<br/>where λ > 0 is a weighting parameter for the relaxed regularizer (cid:107)W(cid:107)2,1 +<br/>that induces row sparsity of the projection matrix W. The meanings of the +<br/>constraints BT B = I,ET E = I,E ≥ 0 are as follows: 1) the orthogonal con- +<br/>straint of B lets each column of B be independent; 2) the orthogonal and +<br/>the nonnegative constraint of E make each row of E has only one non-zero +<br/>element [2]. From 1) and 2), we can clearly interpret B as the basis matrix, +<br/>which has orthogonality and E as the encoding matrix, where the non-zero +<br/>element of each column of ET selects one column in B. +<br/>While optimizing problem (1), T = BET acts like clustering of projected +<br/>data points WT X with orthogonal basis B and encoder E, so T can estimate +<br/>latent cluster centers of the WT X. Then, W successively projects X close +<br/>to corresponding latent cluster centers, which are estimated by T. Note that +<br/>the orthogonal constraint of B makes each projected cluster in WT X be sep- +<br/>arated (independent of each other), and it helps W to be a better projection +<br/>matrix for selecting more discriminative features. If the clustering is directly +<br/>performed on X not on WT X, the orthogonal constraint of B extremely re- +<br/>stricts the degree of freedom of B. However, since features are selected by +<br/>W and the clustering is carried out on WT X in our formulation, so the or- +<br/>thogonal constraint of B is highly reasonable. A schematic illustration of +<br/>the proposed method is shown in Figure 1. +</td></tr><tr><td>47541d04ec24662c0be438531527323d983e958e</td><td>Affective Information Processing +</td></tr><tr><td>474b461cd12c6d1a2fbd67184362631681defa9e</td><td>2014 IEEE International +<br/>Conference on Systems, Man +<br/>and Cybernetics +<br/>(SMC 2014) +<br/>San Diego, California, USA +<br/>5-8 October 2014 +<br/>Pages 1-789 +<br/>IEEE Catalog Number: +<br/>ISBN: +<br/>CFP14SMC-POD +<br/>978-1-4799-3841-4 +<br/>1/5 +</td></tr><tr><td>47d4838087a7ac2b995f3c5eba02ecdd2c28ba14</td><td>JOURNAL OF IEEE TRANSACTIONS ON AFFECTIVE COMPUTING, VOL. XX, NO. X, XXX 2017 +<br/>Automatic Recognition of Facial Displays of +<br/>Unfelt Emotions +<br/>Escalera, Xavier Bar´o, Sylwia Hyniewska, Member, IEEE, J¨uri Allik, +</td></tr><tr><td>47a2727bd60e43f3253247b6d6f63faf2b67c54b</td><td>Semi-supervised Vocabulary-informed Learning +<br/>Disney Research +</td></tr><tr><td>47e3029a3d4cf0a9b0e96252c3dc1f646e750b14</td><td>International Conference on Computer Systems and Technologies - CompSysTech’07 +<br/>Facial Expression Recognition in still pictures and videos using Active +<br/>Appearance Models. A comparison approach. +<br/>Drago(cid:1) Datcu +<br/>Léon Rothkrantz +</td></tr><tr><td>475e16577be1bfc0dd1f74f67bb651abd6d63524</td><td>DAiSEE: Towards User Engagement Recognition in the Wild +<br/>Microsoft +<br/>Vineeth N Balasubramanian +<br/>Indian Institution of Technology Hyderabad +</td></tr><tr><td>471befc1b5167fcfbf5280aa7f908eff0489c72b</td><td>570 +<br/>Class-Specific Kernel-Discriminant +<br/>Analysis for Face Verification +<br/>class problems ( +</td></tr><tr><td>47f8b3b3f249830b6e17888df4810f3d189daac1</td><td></td></tr><tr><td>47e8db3d9adb79a87c8c02b88f432f911eb45dc5</td><td>MAGMA: Multi-level accelerated gradient mirror descent algorithm for +<br/>large-scale convex composite minimization +<br/>July 15, 2016 +</td></tr><tr><td>47aeb3b82f54b5ae8142b4bdda7b614433e69b9a</td><td></td></tr><tr><td>477811ff147f99b21e3c28309abff1304106dbbe</td><td></td></tr><tr><td>47e14fdc6685f0b3800f709c32e005068dfc8d47</td><td></td></tr><tr><td>782188821963304fb78791e01665590f0cd869e8</td><td></td></tr><tr><td>78a4cabf0afc94da123e299df5b32550cd638939</td><td></td></tr><tr><td>78f08cc9f845dc112f892a67e279a8366663e26d</td><td>TECHNISCHE UNIVERSIT ¨AT M ¨UNCHEN +<br/>Lehrstuhl f¨ur Mensch-Maschine-Kommunikation +<br/>Semi-Autonomous Data Enrichment and +<br/>Optimisation for Intelligent Speech Analysis +<br/>Vollst¨andiger Abdruck der von der Fakult¨at f¨ur Elektrotechnik und Informationstechnik +<br/>der Technischen Universit¨at M¨unchen zur Erlangung des akademischen Grades eines +<br/>Doktor-Ingenieurs (Dr.-Ing.) +<br/>genehmigten Dissertation. +<br/>Vorsitzender: +<br/>Univ.-Prof. Dr.-Ing. habil. Dr. h.c. Alexander W. Koch +<br/>Pr¨ufer der Dissertation: +<br/>1. +<br/>Univ.-Prof. Dr.-Ing. habil. Bj¨orn W. Schuller, +<br/>Universit¨at Passau +<br/>2. Univ.-Prof. Gordon Cheng, Ph.D. +<br/>Die Dissertation wurde am 30.09.2014 bei der Technischen Universit¨at M¨unchen einge- +<br/>reicht und durch die Fakult¨at f¨ur Elektrotechnik und Informationstechnik am 07.04.2015 +<br/>angenommen. +</td></tr><tr><td>783f3fccde99931bb900dce91357a6268afecc52</td><td>Hindawi Publishing Corporation +<br/>EURASIP Journal on Image and Video Processing +<br/>Volume 2009, Article ID 945717, 14 pages +<br/>doi:10.1155/2009/945717 +<br/>Research Article +<br/>Adapted Active Appearance Models +<br/>1 SUP ´ELEC/IETR, Avenue de la Boulaie, 35511 Cesson-S´evign´e, France +<br/>2 Orange Labs—TECH/IRIS, 4 rue du clos courtel, 35 512 Cesson S´evign´e, France +<br/>Received 5 January 2009; Revised 2 September 2009; Accepted 20 October 2009 +<br/>Recommended by Kenneth M. Lam +<br/>Active Appearance Models (AAMs) are able to align efficiently known faces under duress, when face pose and illumination are +<br/>controlled. We propose Adapted Active Appearance Models to align unknown faces in unknown poses and illuminations. Our +<br/>proposal is based on the one hand on a specific transformation of the active model texture in an oriented map, which changes the +<br/>AAM normalization process; on the other hand on the research made in a set of different precomputed models related to the most +<br/>adapted AAM for an unknown face. Tests on public and private databases show the interest of our approach. It becomes possible +<br/>to align unknown faces in real-time situations, in which light and pose are not controlled. +<br/>Copyright © 2009 Renaud S´eguier et al. This is an open access article distributed under the Creative Commons Attribution +<br/>License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly +<br/>cited. +<br/>1. Introduction +<br/>All applications related to face analysis and synthesis (Man- +<br/>Machine Interaction, compression in video communication, +<br/>augmented reality) need to detect and then to align the user’s +<br/>face. This latest process consists in the precise localization of +<br/>the eyes, nose, and mouth gravity center. Face detection can +<br/>now be realized in real time and in a rather efficient manner +<br/>[1, 2]; the technical bottleneck lies now in the face alignment +<br/>when it is done in real conditions, which is precisely the +<br/>object of this paper. +<br/>Since such Active Appearance Models (AAMs) as those +<br/>described in [3] exist, it is therefore possible to align faces +<br/>in real time. The AAMs exploit a set of face examples in +<br/>order to extract a statistical model. To align an unknown +<br/>face in new image, the models parameters must be tuned, in +<br/>order to match the analyzed face features in the best possible +<br/>way. There is no difficulty to align a face featuring the same +<br/>characteristics (same morphology, illumination, and pose) +<br/>as those constituting the example data set. Unfortunately, +<br/>AAMs are less outstanding when illumination, pose, and +<br/>face type changes. We suggest in this paper a robust Active +<br/>Appearance Model allowing a real-time implementation. In +<br/>the next section, we will survey the different techniques, +<br/>which aim to increase the AAM robustness. We will see +<br/>that none of them address at the same time the three types +<br/>of robustness, we are interested in pose, illumination, and +<br/>identity. It must be pointed out that we do not consider the +<br/>robustness against occlusion as [4] does, for example, when +<br/>a person moves his hand around the face. +<br/>After a quick introduction of the Active Appearance +<br/>Models and their limitations (Section 3), we will present our +<br/>two main contributions in Section 4.1 in order to improve +<br/>AAM robustness in illumination, pose, and identity. Exper- +<br/>iments will be conducted and discussed in Section 5 before +<br/>drawing a conclusion, suggesting new research directions in +<br/>the last section. +<br/>2. State of the Art +<br/>We propose to classify the methods which lead to an increase +<br/>of the AAM robustness as follows. The specific types of +<br/>dedicated robustness are in italic. +<br/>(i) Preprocess +<br/>(1) Invariant features (illumination) +<br/>(2) Canonical representation (illumination) +<br/>(ii) Parameter space extension +<br/>(1) Light modeling (illumination) +<br/>(2) 3D modeling (pose) +</td></tr><tr><td>7897c8a9361b427f7b07249d21eb9315db189496</td><td></td></tr><tr><td>78f438ed17f08bfe71dfb205ac447ce0561250c6</td><td></td></tr><tr><td>78a11b7d2d7e1b19d92d2afd51bd3624eca86c3c</td><td></td></tr><tr><td>781c2553c4ed2a3147bbf78ad57ef9d0aeb6c7ed</td><td>Int J Comput Vis +<br/>DOI 10.1007/s11263-017-1023-9 +<br/>Tubelets: Unsupervised Action Proposals from Spatiotemporal +<br/>Super-Voxels +<br/>Cees G. M. Snoek1 +<br/>Received: 25 June 2016 / Accepted: 18 May 2017 +<br/>© The Author(s) 2017. This article is an open access publication +</td></tr><tr><td>78df7d3fdd5c32f037fb5cc2a7c104ac1743d74e</td><td>TEMPORAL PYRAMID POOLING CNN FOR ACTION RECOGNITION +<br/>Temporal Pyramid Pooling Based Convolutional +<br/>Neural Network for Action Recognition +</td></tr><tr><td>78fdf2b98cf6380623b0e20b0005a452e736181e</td><td></td></tr><tr><td>788a7b59ea72e23ef4f86dc9abb4450efefeca41</td><td></td></tr><tr><td>8b7191a2b8ab3ba97423b979da6ffc39cb53f46b</td><td>Search Pruning in Video Surveillance Systems: Efficiency-Reliability Tradeoff +<br/>EURECOM +<br/>Sophia Antipolis, France +</td></tr><tr><td>8b8728edc536020bc4871dc66b26a191f6658f7c</td><td></td></tr><tr><td>8b744786137cf6be766778344d9f13abf4ec0683</td><td>978-1-4799-9988-0/16/$31.00 ©2016 IEEE +<br/>2697 +<br/>ICASSP 2016 +</td></tr><tr><td>8bf647fed40bdc9e35560021636dfb892a46720e</td><td>Learning to Hash-tag Videos with Tag2Vec +<br/>CVIT, KCIS, IIIT Hyderabad, India +<br/>P J Narayanan +<br/>http://cvit.iiit.ac.in/research/projects/tag2vec +<br/>Figure 1. Learning a direct mapping from videos to hash-tags : sample frames from short video clips with user-given hash-tags +<br/>(left); a sample frame from a query video and hash-tags suggested by our system for this query (right). +</td></tr><tr><td>8bb21b1f8d6952d77cae95b4e0b8964c9e0201b0</td><td>Methoden +<br/>at 11/2013 +<br/>(cid:2)(cid:2)(cid:2) +<br/>Multimodale Interaktion +<br/>auf einer sozialen Roboterplattform +<br/>Multimodal Interaction on a Social Robotic Platform +<br/>Zusammenfassung Dieser Beitrag beschreibt die multimo- +<br/>dalen Interaktionsmöglichkeiten mit der Forschungsroboter- +<br/>plattform ELIAS. Zunächst wird ein Überblick über die Ro- +<br/>boterplattform sowie die entwickelten Verarbeitungskompo- +<br/>nenten gegeben, die Einteilung dieser Komponenten erfolgt +<br/>nach dem Konzept von wahrnehmenden und agierenden Mo- +<br/>dalitäten. Anschließend wird das Zusammenspiel der Kom- +<br/>ponenten in einem multimodalen Spieleszenario näher be- +<br/>trachtet. (cid:2)(cid:2)(cid:2) Summary +<br/>This paper presents the mul- +<br/>timodal +<br/>interaction capabilities of the robotic research plat- +<br/>form ELIAS. An overview of the robotic platform as well +<br/>as the developed processing components is presented, the +<br/>classification of the components follows the concept of sen- +<br/>sing and acting modalities. Finally, +<br/>the interplay between +<br/>those components within a multimodal gaming scenario is +<br/>described. +<br/>Schlagwörter Mensch-Roboter-Interaktion, Multimodalität, Gesten, Blick (cid:2)(cid:2)(cid:2) Keywords Human-robot interaction, +<br/>multimodal, gestures, gaze +<br/>1 Einleitung +<br/>Eine intuitive und natürliche Bedienbarkeit der zuneh- +<br/>mend komplexeren Technik wird für den Menschen +<br/>immer wichtiger, da im heutigen Alltag eine Vielzahl an +<br/>technischen Geräten mit wachsendem Funktionsumfang +<br/>anzutreffen ist. Unterschiedliche Aktivitäten in der For- +<br/>schungsgemeinschaft haben sich schon seit längerer Zeit +<br/>mit verbalen sowie nonverbalen Kommunikationsformen +<br/>(bspw. Emotions- und Gestenerkennung) in der Mensch- +<br/>Maschine-Interaktion beschäftigt. Gerade in der jüngeren +<br/>Zeit trugen auf diesem Forschungsfeld unterschiedliche +<br/>Innovationen (bspw. Touchscreen, Gestensteuerung im +<br/>Fernseher) dazu bei, dass intuitive und natürliche Bedien- +<br/>konzepte mehr und mehr im Alltag Verwendung finden. +<br/>Auch Möglichkeiten zur Sprach- und Gestensteuerung +<br/>von Konsolen und Mobiltelefonen finden heute vermehr- +<br/>ten Einsatz in der Gerätebedienung. Diese natürlicheren +<br/>und multimodalen Benutzerschnittstellen sind dem Nut- +<br/>zer schnell zugänglich und erlauben eine intuitivere +<br/>Interaktion mit komplexen technischen Geräten. +<br/>Auch für Robotersysteme bietet sich eine multimodale +<br/>Interaktion an, um die Benutzung und den Zugang zu +<br/>den Funktionalitäten zu vereinfachen. Der Mensch soll +<br/>in seiner Kommunikation idealerweise vollkommene Ent- +<br/>scheidungsfreiheit bei der Wahl der Modalitäten haben, +<br/>um sein gewünschtes Ziel zu erreichen. Dafür werden +<br/>in diesem Beitrag die wahrnehmenden und agieren- +<br/>den Modalitäten einer, rein auf Kommunikationsaspekte +<br/>reduzierten, Forschungsroboterplattform beispielhaft in +<br/>einer Spieleanwendung untersucht. +<br/>1.1 Struktur des Beitrags +<br/>In diesem Beitrag wird zunächst ein kurzer Über- +<br/>blick über die multimodale Interaktion im Allgemeinen +<br/>gegeben, hierbei erfolgt eine Betrachtung nach wahr- +<br/>nehmenden und agierenden Modalitäten. Im nächsten +<br/>Abschnitt werden Arbeiten vorgestellt, die sich auch mit +<br/>multimodalen Robotersystemen beschäftigen. Im darauf +<br/>folgenden Abschnitt wird die Roboterplattform ELIAS +<br/>mit den wahrnehmenden, verarbeitenden und agierenden +<br/>at – Automatisierungstechnik 61 (2013) 11 / DOI 10.1515/auto.2013.1062 © Oldenbourg Wissenschaftsverlag +<br/> - 10.1515/auto.2013.1062 +<br/>Downloaded from De Gruyter Online at 09/27/2016 10:08:34PM +<br/>via Technische Universität München +<br/>737 +</td></tr><tr><td>8b1db0894a23c4d6535b5adf28692f795559be90</td><td>Biometric and Surveillance Technology for Human and Activity Identification X, edited by Ioannis Kakadiaris, +<br/>Walter J. Scheirer, Laurence G. Hassebrook, Proc. of SPIE Vol. 8712, 87120Q · © 2013 SPIE +<br/>CCC code: 0277-786X/13/$18 · doi: 10.1117/12.2018974 +<br/>Proc. of SPIE Vol. 8712 87120Q-1 +</td></tr><tr><td>134db6ca13f808a848321d3998e4fe4cdc52fbc2</td><td>IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART B: CYBERNETICS, VOL. 36, NO. 2, APRIL 2006 +<br/>433 +<br/>Dynamics of Facial Expression: Recognition of +<br/>Facial Actions and Their Temporal Segments +<br/>From Face Profile Image Sequences +</td></tr><tr><td>133dd0f23e52c4e7bf254e8849ac6f8b17fcd22d</td><td>This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI +<br/>Active Clustering with Model-Based +<br/>Uncertainty Reduction +</td></tr><tr><td>1369e9f174760ea592a94177dbcab9ed29be1649</td><td>Geometrical Facial Modeling for Emotion Recognition +</td></tr><tr><td>133900a0e7450979c9491951a5f1c2a403a180f0</td><td>JOURNAL OF LATEX CLASS FILES +<br/>Social Grouping for Multi-target Tracking and +<br/>Head Pose Estimation in Video +</td></tr><tr><td>13141284f1a7e1fe255f5c2b22c09e32f0a4d465</td><td>Object Tracking by +<br/>Oversampling Local Features +</td></tr><tr><td>133da0d8c7719a219537f4a11c915bf74c320da7</td><td>International Journal of Computer Applications (0975 – 8887) +<br/>Volume 123 – No.4, August 2015 +<br/>A Novel Method for 3D Image Segmentation with Fusion +<br/>of Two Images using Color K-means Algorithm +<br/>Dept. of CSE +<br/>ITM Universe +<br/>Gwalior +<br/>Dept. of CSE +<br/>ITM Universe +<br/>Gwalior +<br/>two +</td></tr><tr><td>133f01aec1534604d184d56de866a4bd531dac87</td><td>Effective Unconstrained Face Recognition by +<br/>Combining Multiple Descriptors and Learned +<br/>Background Statistics +</td></tr><tr><td>13841d54c55bd74964d877b4b517fa94650d9b65</td><td>Generalised Ambient Reflection Models for Lambertian and +<br/>Phong Surfaces +<br/>Author +<br/>Zhang, Paul, Gao, Yongsheng +<br/>Published +<br/>2009 +<br/>Conference Title +<br/>Proceedings of the 2009 IEEE International Conference on Image Processing (ICIP 2009) +<br/>DOI +<br/>https://doi.org/10.1109/ICIP.2009.5413812 +<br/>Copyright Statement +<br/>© 2009 IEEE. Personal use of this material is permitted. However, permission to reprint/ +<br/>republish this material for advertising or promotional purposes or for creating new collective +<br/>works for resale or redistribution to servers or lists, or to reuse any copyrighted component of +<br/>this work in other works must be obtained from the IEEE. +<br/>Downloaded from +<br/>http://hdl.handle.net/10072/30001 +<br/>Griffith Research Online +<br/>https://research-repository.griffith.edu.au +</td></tr><tr><td>131e395c94999c55c53afead65d81be61cd349a4</td><td></td></tr><tr><td>1384a83e557b96883a6bffdb8433517ec52d0bea</td><td></td></tr><tr><td>13fd0a4d06f30a665fc0f6938cea6572f3b496f7</td><td></td></tr><tr><td>13afc4f8d08f766479577db2083f9632544c7ea6</td><td>Multiple Kernel Learning for +<br/>Emotion Recognition in the Wild +<br/>Machine Perception Laboratory +<br/>UCSD +<br/>EmotiW Challenge, ICMI, 2013 +<br/>1 +</td></tr><tr><td>13d9da779138af990d761ef84556e3e5c1e0eb94</td><td>Int J Comput Vis (2008) 77: 3–24 +<br/>DOI 10.1007/s11263-007-0093-5 +<br/>Learning to Locate Informative Features for Visual Identification +<br/>Received: 18 August 2005 / Accepted: 11 September 2007 / Published online: 9 November 2007 +<br/>© Springer Science+Business Media, LLC 2007 +</td></tr><tr><td>7f533bd8f32525e2934a66a5b57d9143d7a89ee1</td><td>Audio-Visual Identity Grounding for Enabling Cross Media Search +<br/>Paper ID 22 +</td></tr><tr><td>7f44f8a5fd48b2d70cc2f344b4d1e7095f4f1fe5</td><td>Int J Comput Vis (2016) 119:60–75 +<br/>DOI 10.1007/s11263-015-0839-4 +<br/>Sparse Output Coding for Scalable Visual Recognition +<br/>Received: 15 May 2013 / Accepted: 16 June 2015 / Published online: 26 June 2015 +<br/>© Springer Science+Business Media New York 2015 +</td></tr><tr><td>7f4bc8883c3b9872408cc391bcd294017848d0cf</td><td> +<br/> +<br/>Computer +<br/>Sciences +<br/>Department +<br/>The Multimodal Focused Attribute Model: A Nonparametric +<br/>Bayesian Approach to Simultaneous Object Classification and +<br/>Attribute Discovery +<br/>Technical Report #1697 +<br/>January 2012 +<br/> +</td></tr><tr><td>7f6061c83dc36633911e4d726a497cdc1f31e58a</td><td>YouTube-8M: A Large-Scale Video Classification +<br/>Benchmark +<br/>Paul Natsev +<br/>Google Research +</td></tr><tr><td>7f36dd9ead29649ed389306790faf3b390dc0aa2</td><td>MOVEMENT DIFFERENCES BETWEEN DELIBERATE +<br/>AND SPONTANEOUS FACIAL EXPRESSIONS: +<br/>ZYGOMATICUS MAJOR ACTION IN SMILING +</td></tr><tr><td>7f6cd03e3b7b63fca7170e317b3bb072ec9889e0</td><td>A Face Recognition Signature Combining Patch-based +<br/>Features with Soft Facial Attributes +<br/>L. Zhang, P. Dou, I.A. Kakadiaris +<br/>Computational Biomedicine Lab, 4849 Calhoun Rd, Rm 373, Houston, TX 77204 +</td></tr><tr><td>7f3a73babe733520112c0199ff8d26ddfc7038a0</td><td></td></tr><tr><td>7f205b9fca7e66ac80758c4d6caabe148deb8581</td><td>Page 1 of 47 +<br/>Computing Surveys +<br/>A Survey on Mobile Social Signal Processing +<br/>Understanding human behaviour in an automatic but non-intrusive manner is an important area for various applications. This requires the +<br/>collaboration of information technology with human sciences to transfer existing knowledge of human behaviour into self-acting tools. These +<br/>tools will reduce human error that is introduced by current obtrusive methods such as questionnaires. To achieve unobtrusiveness, we focus on +<br/>exploiting the pervasive and ubiquitous character of mobile devices. +<br/>In this article, a survey of existing techniques for extracting social behaviour through mobile devices is provided. Initially we expose the +<br/>terminology used in the area and introduce a concrete architecture for social signal processing applications on mobile phones, constituted by +<br/>sensing, social interaction detection, behavioural cues extraction, social signal inference and social behaviour understanding. Furthermore, we +<br/>present state-of-the-art techniques applied to each stage of the process. Finally, potential applications are shown while arguing about the main +<br/>challenges of the area. +<br/>Categories and Subject Descriptors: General and reference [Document Types]: Surveys and Overviews; Human-centered computing [Collab- +<br/>orative and social computing, Ubiquitous and mobile computing] +<br/>General Terms: Design, Theory, Human Factors, Performance +<br/>Additional Key Words and Phrases: Social Signal Processing, mobile phones, social behaviour +<br/>ACM Reference Format: +<br/>Processing. ACM V, N, Article A (January YYYY), 35 pages. +<br/>DOI:http://dx.doi.org/10.1145/0000000.0000000 +<br/>1. INTRODUCTION +<br/>Human behaviour understanding has received a great deal of interest since the beginning of the previous century. +<br/>People initially conducted research on the way animals behave when they are surrounded by creatures of the same +<br/>species. Acquiring basic underlying knowledge of animal relations led to extending this information to humans +<br/>in order to understand social behaviour, social relations etc. Initial experiments were conducted by empirically +<br/>observing people and retrieving feedback from them. These methods gave rise to well-established psychological +<br/>approaches for understanding human behaviour, such as surveys, questionnaires, camera recordings and human +<br/>observers. Nevertheless, these methods introduce several limitations including various sources of error. Complet- +<br/>ing surveys and questionnaires induces partiality, unconcern etc. [Groves 2004], human error [Reason 1990], and +<br/>additional restrictions in scalability of the experiments. Accumulating these research problems leads to a common +<br/>challenge, the lack of automation in an unobtrusive manner. +<br/>An area that has focussed on detecting social behaviour automatically and has received a great amount of at- +<br/>tention is Social Signal Processing (SSP). The main target of the field is to model, analyse and synthesise human +<br/>behaviour with limited user intervention. To achieve these targets, researchers presented three key terms which +</td></tr><tr><td>7a9ef21a7f59a47ce53b1dff2dd49a8289bb5098</td><td></td></tr><tr><td>7af38f6dcfbe1cd89f2307776bcaa09c54c30a8b</td><td>eaig i C e Vii ad Beyd: +<br/>Deve +<br/>h . Weg +<br/>Deae f C e Sciece +<br/>ichiga Sae Uiveiy +<br/>Ea aig 48824 +<br/>Abac +<br/>Thi chae id ce wha i ca +<br/>aic +<br/>ve +<br/>by h a cgiive deve +<br/>ih i deeied befe he \bih" f he ye. Afe he \bih" i eab +<br/> +<br/>aach i ea +<br/>deve +<br/>way aia +<br/> ea whi +<br/>de deve +</td></tr><tr><td>7a81967598c2c0b3b3771c1af943efb1defd4482</td><td>Do We Need More Training Data? +</td></tr><tr><td>7ad77b6e727795a12fdacd1f328f4f904471233f</td><td>Supervised Local Descriptor Learning +<br/>for Human Action Recognition +</td></tr><tr><td>7a97de9460d679efa5a5b4c6f0b0a5ef68b56b3b</td><td></td></tr><tr><td>7aa4c16a8e1481629f16167dea313fe9256abb42</td><td>978-1-5090-4117-6/17/$31.00 ©2017 IEEE +<br/>2981 +<br/>ICASSP 2017 +</td></tr><tr><td>7a85b3ab0efb6b6fcb034ce13145156ee9d10598</td><td></td></tr><tr><td>7ab930146f4b5946ec59459f8473c700bcc89233</td><td></td></tr><tr><td>7ad7897740e701eae455457ea74ac10f8b307bed</td><td>Random Subspace Two-dimensional LDA for Face Recognition* +</td></tr><tr><td>7a7b1352d97913ba7b5d9318d4c3d0d53d6fb697</td><td>Attend and Rectify: a Gated Attention +<br/>Mechanism for Fine-Grained Recovery +<br/>†Computer Vision Center and Universitat Aut`onoma de Barcelona (UAB), +<br/>Campus UAB, 08193 Bellaterra, Catalonia Spain +<br/>‡Visual Tagging Services, Parc de Recerca, Campus UAB +</td></tr><tr><td>1451e7b11e66c86104f9391b80d9fb422fb11c01</td><td>IET Signal Processing +<br/>Research Article +<br/>Image privacy protection with secure JPEG +<br/>transmorphing +<br/>ISSN 1751-9675 +<br/>Received on 30th December 2016 +<br/>Revised 13th July 2017 +<br/>Accepted on 11th August 2017 +<br/>doi: 10.1049/iet-spr.2016.0756 +<br/>www.ietdl.org +<br/>1Multimedia Signal Processing Group, Electrical Engineering Department, EPFL, Station 11, Lausanne, Switzerland +</td></tr><tr><td>14761b89152aa1fc280a33ea4d77b723df4e3864</td><td></td></tr><tr><td>14fa27234fa2112014eda23da16af606db7f3637</td><td></td></tr><tr><td>1459d4d16088379c3748322ab0835f50300d9a38</td><td>Cross-Domain Visual Matching via Generalized +<br/>Similarity Measure and Feature Learning +</td></tr><tr><td>14e949f5754f9e5160e8bfa3f1364dd92c2bb8d6</td><td></td></tr><tr><td>1450296fb936d666f2f11454cc8f0108e2306741</td><td>Learning to Discover Cross-Domain Relations +<br/>with Generative Adversarial Networks +</td></tr><tr><td>14fdce01c958043140e3af0a7f274517b235adf3</td><td></td></tr><tr><td>141eab5f7e164e4ef40dd7bc19df9c31bd200c5e</td><td></td></tr><tr><td>14e759cb019aaf812d6ac049fde54f40c4ed1468</td><td>Subspace Methods +<br/>Synonyms +<br/>{ Multiple similarity method +<br/>Related Concepts +<br/>{ Principal component analysis (PCA) +<br/>{ Subspace analysis +<br/>{ Dimensionality reduction +<br/>De(cid:12)nition +<br/>Subspace analysis in computer vision is a generic name to describe a general +<br/>framework for comparison and classification of subspaces. A typical approach in +<br/>subspace analysis is the subspace method (SM) that classify an input pattern +<br/>vector into several classes based on the minimum distance or angle between the +<br/>input pattern vector and each class subspace, where a class subspace corresponds +<br/>to the distribution of pattern vectors of the class in high dimensional vector +<br/>space. +<br/>Background +<br/>Comparison and classification of subspaces has been one of the central prob- +<br/>lems in computer vision, where an image set of an object to be classified is +<br/>compactly represented by a subspace in high dimensional vector space. +<br/>The subspace method is one of the most effective classification method in +<br/>subspace analysis, which was developed by two Japanese researchers, Watanabe +<br/>and Iijima around 1970, independently [1, 2]. Watanabe and Iijima named their +<br/>methods the CLAFIC [3] and the multiple similarity method [4], respectively. +<br/>The concept of the subspace method is derived from the observation that pat- +<br/>terns belonging to a class forms a compact cluster in high dimensional vector +<br/>space, where, for example, a w×h pixels image pattern is usually represented as a +<br/>vector in w×h-dimensional vector space. The compact cluster can be represented +<br/>by a subspace, which is generated by using Karhunen-Lo`eve (KL) expansion, also +<br/>known as the principal component analysis (PCA). Note that a subspace is gen- +<br/>erated for each class, unlike the Eigenface Method [5] in which only one subspace +<br/>(called eigenspace) is generated. +<br/>The SM has been known as one of the most useful methods in pattern recog- +<br/>nition field, since its algorithm is very simple and it can handle classification +<br/>of multiple classes. However, its classification performance was not sufficient for +<br/>many applications in practice, because class subspaces are generated indepen- +<br/>dently of each other [1]. There is no reason to assume a priori that each class +</td></tr><tr><td>148eb413bede35487198ce7851997bf8721ea2d6</td><td>People Search in Surveillance Videos +<br/>Four Eyes Lab, UCSB +<br/>IBM Research +<br/>IBM Research +<br/>IBM Research +<br/>Four Eyes Lab, UCSB +<br/>INTRODUCTION +<br/>1. +<br/>In traditional surveillance scenarios, users are required to +<br/>watch video footage corresponding to extended periods of +<br/>time in order to find events of interest. However, this pro- +<br/>cess is resource-consuming, and suffers from high costs of +<br/>employing security personnel. The field of intelligent vi- +<br/>sual surveillance [2] seeks to address these issues by applying +<br/>computer vision techniques to automatically detect specific +<br/>events in long video streams. The events can then be pre- +<br/>sented to the user or be indexed into a database to allow +<br/>queries such as “show me the red cars that entered a given +<br/>parking lot from 7pm to 9pm on Monday” or “show me the +<br/>faces of people who left the city’s train station last week.” +<br/>In this work, we are interested in analyzing people, by ex- +<br/>tracting information that can be used to search for them in +<br/>surveillance videos. Current research on this topic focuses +<br/>on approaches based on face recognition, where the goal is +<br/>to establish the identity of a person given an image of a +<br/>face. However, face recognition is still a very challenging +<br/>problem, especially in low resolution images with variations +<br/>in pose and lighting, which is often the case in surveillance +<br/>data. State-of-the-art face recognition systems [1] require +<br/>a fair amount of resolution in order to produce reliable re- +<br/>sults, but in many cases this level of detail is not available +<br/>in surveillance applications. +<br/>We approach the problem in an alternative way, by avoiding +<br/>face recognition and proposing a framework for finding peo- +<br/>ple based on parsing the human body and exploiting part +<br/>attributes. Those include visual attributes such as facial hair +<br/>type (beards, mustaches, absence of facial hair), type of eye- +<br/>wear (sunglasses, eyeglasses, absence of glasses), hair type +<br/>(baldness, hair, wearing a hat), and clothing color. While +<br/>face recognition is still a difficult problem, accurate and ef- +<br/>ficient face detectors1 based on learning approaches [6] are +<br/>available. Those have been demonstrated to work well on +<br/>challenging low-resolution images, with variations in pose +<br/>and lighting. In our method, we employ this technology to +<br/>design detectors for facial attributes from large sets of train- +<br/>ing data. +<br/>1The face detection problem consists of localizing faces in +<br/>images, while face recognition aims to establish the identity +<br/>of a person given an image of a face. Face detection is a +<br/>challenging problem, but it is arguably not as complex as +<br/>face recognition. +<br/>Our technique falls into the category of short term recogni- +<br/>tion methods, taking advantage of features present in brief +<br/>intervals in time, such as clothing color, hairstyle, and makeup, +<br/>which are generally considered an annoyance in face recogni- +<br/>tion methods. There are several applications that naturally +<br/>fit within a short term recognition framework. An example +<br/>is in criminal investigation, when the police are interested in +<br/>locating a suspect. In those cases, eyewitnesses typically fill +<br/>out a suspect description form, where they indicate personal +<br/>traits of the suspect as seen at the moment when the crime +<br/>was committed. Those include facial hair type, hair color, +<br/>clothing type, etc. Based on that description, the police +<br/>manually scan the entire video archive looking for a person +<br/>with similar characteristics. This process is tedious and time +<br/>consuming, and could be drastically accelerated by the use +<br/>of our technique. Another application is on finding missing +<br/>people. Parents looking for their children in an amusement +<br/>park could provide a description including clothing and eye- +<br/>wear type, and videos from multiple cameras in the park +<br/>would then be automatically searched. +</td></tr><tr><td>1473a233465ea664031d985e10e21de927314c94</td><td></td></tr><tr><td>140c95e53c619eac594d70f6369f518adfea12ef</td><td>Pushing the Frontiers of Unconstrained Face Detection and Recognition: IARPA Janus Benchmark A +<br/>The development of accurate and scalable unconstrained face recogni- +<br/>tion algorithms is a long term goal of the biometrics and computer vision +<br/>communities. The term “unconstrained” implies a system can perform suc- +<br/>cessful identifications regardless of face image capture presentation (illumi- +<br/>nation, sensor, compression) or subject conditions (facial pose, expression, +<br/>occlusion). While automatic, as well as human, face identification in certain +<br/>scenarios may forever be elusive, such as when a face is heavily occluded or +<br/>captured at very low resolutions, there still remains a large gap between au- +<br/>tomated systems and human performance on familiar faces. In order to close +<br/>this gap, large annotated sets of imagery are needed that are representative +<br/>of the end goals of unconstrained face recognition. This will help continue +<br/>to push the frontiers of unconstrained face detection and recognition, which +<br/>are the primary goals of the IARPA Janus program. +<br/>The current state of the art in unconstrained face recognition is high +<br/>accuracy (roughly 99% true accept rate at a false accept rate of 1.0%) on +<br/>faces that can be detected with a commodity face detectors, but unknown +<br/>accuracy on other faces. Despite the fact that face detection and recognition +<br/>research generally has advanced somewhat independently, the frontal face +<br/>detector filtering approach used for key in the wild face recognition datasets +<br/>means that progress in face recognition is currently hampered by progress +<br/>in face detection. Hence, a major need exists for a face recognition dataset +<br/>that captures as wide of a range of variations as possible to offer challenges +<br/>to both face detection as well as face recognition. +<br/>In this paper we introduce the IARPA Janus Benchmark A (IJB-A), +<br/>which is publicly available for download. The IJB-A contains images and +<br/>videos from 500 subjects captured from “in the wild” environment. All la- +<br/>belled subjects have been manually localized with bounding boxes for face +<br/>detection, as well as fiducial landmarks for the center of the two eyes (if +<br/>visible) and base of the nose. Manual bounding box annotations for all non- +<br/>labelled subjects (i.e., other persons captured in the imagery) have been cap- +<br/>tured as well. All imagery is Creative Commons licensed, which is a license +<br/>that allows open re-distribution provided proper attribution is made to the +<br/>data creator. The subjects have been intentionally sampled to contain wider +<br/>geographic distribution than previous datasets. Recognition and detection +<br/>protocols are provided which are motivated by operational deployments of +<br/>face recognition systems. An example of images and video from IJB-A can +<br/>be found in Figure 3. +<br/>The IJB-A dataset has the following claimed contributions: (i) The most +<br/>unconstrained database released to date; (ii) The first joint face detection and +<br/>face recognition benchmark dataset collected in the wild; (iii) Meta-data +<br/>providing subject gender and skin color, and occlusion (eyes, mouth/nose, +<br/>and forehead), facial hear, and coarse pose information for each imagery +<br/>instance; (iv) Widest geographic distribution of any public face dataset; (v) +<br/>The first in the wild dataset to contain a mixture of images and videos; (vi) +<br/>Clear authority for re-distribution; (vii) Protocols for identification (search) +<br/>and verification (compare); (viii) Baseline accuracies from off the shelf de- +<br/>tectors and recognition algorithms; and (ix) Protocols for both template and +<br/>model-based face recognition. +<br/>Every subject in the dataset contains at least five images and one video. +<br/>IJB-A consists of a total of 5,712 images and 2,085 videos, with an average +<br/>of 11.4 images and 4.2 videos per subject. +</td></tr><tr><td>142dcfc3c62b1f30a13f1f49c608be3e62033042</td><td>Adaptive Region Pooling for Object Detection +<br/>UC Merced +<br/>Qualcomm Research, San Diego +<br/>UC Merced +</td></tr><tr><td>14e428f2ff3dc5cf96e5742eedb156c1ea12ece1</td><td>Facial Expression Recognition Using Neural Network Trained with Zernike +<br/>Moments +<br/>Dept. Génie-Electrique +<br/>Université M.C.M Souk-Ahras +<br/>Souk-Ahras, Algeria +</td></tr><tr><td>14a5feadd4209d21fa308e7a942967ea7c13b7b6</td><td>978-1-4673-0046-9/12/$26.00 ©2012 IEEE +<br/>1025 +<br/>ICASSP 2012 +</td></tr><tr><td>14fee990a372bcc4cb6dc024ab7fc4ecf09dba2b</td><td>Modeling Spatio-Temporal Human Track Structure for Action +<br/>Localization +</td></tr><tr><td>14ee4948be56caeb30aa3b94968ce663e7496ce4</td><td>Jang, Y; Gunes, H; Patras, I +<br/>© Copyright 2018 IEEE +<br/>For additional information about this publication click this link. +<br/>http://qmro.qmul.ac.uk/xmlui/handle/123456789/36405 +<br/>Information about this research object was correct at the time of download; we occasionally +<br/>make corrections to records, please therefore check the published record when citing. For +</td></tr><tr><td>8ee62f7d59aa949b4a943453824e03f4ce19e500</td><td>Robust Head-Pose Estimation Based on +<br/>Partially-Latent Mixture of Linear Regression +<br/>∗INRIA Grenoble Rhˆone-Alpes, Montbonnot Saint-Martin, France +<br/>†INRIA Rennes Bretagne Atlantique, Rennes, France +</td></tr><tr><td>8e33183a0ed7141aa4fa9d87ef3be334727c76c0</td><td>– COS429 Written Report, Fall 2017 – +<br/>Robustness of Face Recognition to Image Manipulations +<br/>1. Motivation +<br/>We can often recognize pictures of people we know even if the image has low resolution or obscures +<br/>part of the face, if the camera angle resulted in a distorted image of the subject’s face, or if the +<br/>subject has aged or put on makeup since we last saw them. Although this is a simple recognition task +<br/>for a human, when we think about how we accomplish this task, it seems non-trivial for computer +<br/>algorithms to recognize faces despite visual changes. +<br/>Computer facial recognition is relied upon for many application where accuracy is important. +<br/>Facial recognition systems have applications ranging from airport security and suspect identification +<br/>to personal device authentication and face tagging [7]. In these real-world applications, the system +<br/>must continue to recognize images of a person who looks slightly different due to the passage of +<br/>time, a change in environment, or a difference in clothing. +<br/>Therefore, we are interested in investigating face recognition algorithms and their robustness to +<br/>image changes resulting from realistically plausible manipulations. Furthermore, we are curious +<br/>about whether the impact of image manipulations on computer algorithms’ face recognition ability +<br/>mirrors related insights from neuroscience about humans’ face recognition abilities. +<br/>2. Goal +<br/>In this project, we implement both face recognition algorithms and image manipulations. We then +<br/>analyze the impact of each image manipulation on the recognition accuracy each algorithm, and +<br/>how these influences depend on the accuracy of each algorithm on non-manipulated images. +<br/>3. Background and Related Work +<br/>Researchers have developed a wide variety of face recognition algorithms, such as traditional +<br/>statistical methods such as PCA, more opaque methods such as deep neural networks, and proprietary +<br/>systems used by governments and corporations [1][13][14]. +<br/>Similarly, others have developed image manipulations using principles from linear algebra, such +<br/>as mimicking distortions from lens distortions, as well as using neural networks, such as a system +<br/>for transforming images according to specified characteristics [12][16]. +<br/>Furthermore, researchers in psychology have studied face recognition in humans. A study of +<br/>“super-recognizers” (people with extraordinarily high powers of face recognition) and “developmen- +<br/>tal prosopagnosics” (people with severely impaired face recognition abilities) found that inverting +<br/>images of faces impaired recognition ability more for people with stronger face recognition abilities +<br/>[11]. This could indicate that image manipulations tend to equalize face recognition abilities, and +<br/>we investigate whether this is the case with the manipulations and face recognition algorithms we +<br/>test. +</td></tr><tr><td>8e3d0b401dec8818cd0245c540c6bc032f169a1d</td><td>McGan: Mean and Covariance Feature Matching GAN +</td></tr><tr><td>8e94ed0d7606408a0833e69c3185d6dcbe22bbbe</td><td>© 2012 IEEE. Personal use of this material is permitted. Permission from IEEE +<br/>must be obtained for all other uses, in any current or future media, including +<br/>reprinting/republishing this material for advertising or promotional purposes, +<br/>creating new collective works, for resale or redistribution to servers or lists, or +<br/>reuse of any copyrighted component of this work in other works. +<br/>Pre-print of article that will appear at WACV 2012. +</td></tr><tr><td>8e461978359b056d1b4770508e7a567dbed49776</td><td>LOMo: Latent Ordinal Model for Facial Analysis in Videos +<br/>Marian Bartlett1,∗,‡ +<br/>1UCSD, USA +<br/>2MPI for Informatics, Germany +<br/>3IIT Kanpur, India +</td></tr><tr><td>8ea30ade85880b94b74b56a9bac013585cb4c34b</td><td>FROM TURBO HIDDEN MARKOV MODELS TO TURBO STATE-SPACE MODELS +<br/>Institut Eur´ecom +<br/>Multimedia Communications Department +<br/>BP 193, 06904 Sophia Antipolis Cedex, France +</td></tr><tr><td>8ed32c8fad924736ebc6d99c5c319312ba1fa80b</td><td></td></tr><tr><td>8e8e3f2e66494b9b6782fb9e3f52aeb8e1b0d125</td><td>in any current or +<br/>future media, +<br/>for all other uses, +<br/> 2012 IEEE. Personal use of this material is permitted. Permission from IEEE must be +<br/>obtained +<br/>including +<br/>reprinting/republishing this material for advertising or promotional purposes, creating +<br/>new collective works, for resale or redistribution to servers or lists, or reuse of any +<br/>copyrighted component of this work in other works. +<br/>Pre-print of article that will appear at BTAS 2012.!! +</td></tr><tr><td>8e378ef01171b33c59c17ff5798f30293fe30686</td><td>Lehrstuhl f¨ur Mensch-Maschine-Kommunikation +<br/>der Technischen Universit¨at M¨unchen +<br/>A System for Automatic Face Analysis +<br/>Based on +<br/>Statistical Shape and Texture Models +<br/>Ronald M¨uller +<br/>Vollst¨andiger Abdruck der von der Fakult¨at +<br/>f¨ur Elektrotechnik und Informationstechnik +<br/>der Technischen Universit¨at M¨unchen +<br/>zur Erlangung des akademischen Grades eines +<br/>Doktor-Ingenieurs +<br/>genehmigten Dissertation +<br/>Vorsitzender: Prof. Dr. rer. nat. Bernhard Wolf +<br/>Pr¨ufer der Dissertation: +<br/>1. Prof. Dr.-Ing. habil. Gerhard Rigoll +<br/>2. Prof. Dr.-Ing. habil. Alexander W. Koch +<br/>Die Dissertation wurde am 28.02.2008 bei der Technischen Universit¨at M¨unchen +<br/>eingereicht und durch die Fakult¨at f¨ur Elektrotechnik und Informationstechnik +<br/>am 18.09.2008 angenommen. +</td></tr><tr><td>8ed051be31309a71b75e584bc812b71a0344a019</td><td>Class-based feature matching across unrestricted +<br/>transformations +</td></tr><tr><td>8e36100cb144685c26e46ad034c524b830b8b2f2</td><td>Modeling Facial Geometry using Compositional VAEs +<br/>1 ´Ecole Polytechnique F´ed´erale de Lausanne +<br/>2Facebook Reality Labs, Pittsburgh +</td></tr><tr><td>8e0becfc5fe3ecdd2ac93fabe34634827b21ef2b</td><td>International Journal of Computer Vision manuscript No. +<br/>(will be inserted by the editor) +<br/>Learning from Longitudinal Face Demonstration - +<br/>Where Tractable Deep Modeling Meets Inverse Reinforcement Learning +<br/>Savvides · Tien D. Bui +<br/>Received: date / Accepted: date +</td></tr><tr><td>225fb9181545f8750061c7693661b62d715dc542</td><td></td></tr><tr><td>22043cbd2b70cb8195d8d0500460ddc00ddb1a62</td><td>Separability-Oriented Subclass Discriminant +<br/>Analysis +</td></tr><tr><td>22137ce9c01a8fdebf92ef35407a5a5d18730dde</td><td></td></tr><tr><td>22dada4a7ba85625824489375184ba1c3f7f0c8f</td><td></td></tr><tr><td>223ec77652c268b98c298327d42aacea8f3ce23f</td><td>TR-CS-11-02 +<br/>Acted Facial Expressions In The Wild +<br/>Database +<br/>September 2011 +<br/>ANU Computer Science Technical Report Series +</td></tr><tr><td>227b18fab568472bf14f9665cedfb95ed33e5fce</td><td>Compositional Dictionaries for Domain Adaptive +<br/>Face Recognition +</td></tr><tr><td>227b1a09b942eaf130d1d84cdcabf98921780a22</td><td>Yang et al. EURASIP Journal on Advances in Signal Processing (2018) 2018:51 +<br/>https://doi.org/10.1186/s13634-018-0572-6 +<br/>EURASIP Journal on Advances +<br/>in Signal Processing +<br/>R ES EAR CH +<br/>Multi-feature shape regression for face +<br/>alignment +<br/>Open Access +</td></tr><tr><td>22dabd4f092e7f3bdaf352edd925ecc59821e168</td><td> Deakin Research Online +<br/>This is the published version: +<br/>An, Senjian, Liu, Wanquan and Venkatesh, Svetha 2008, Exploiting side information in +<br/>locality preserving projection, in CVPR 2008 : Proceedings of the 26th IEEE Conference on +<br/>Computer Vision and Pattern Recognition, IEEE, Washington, D. C., pp. 1-8. +<br/>Available from Deakin Research Online: +<br/>http://hdl.handle.net/10536/DRO/DU:30044576 +<br/> +<br/>Reproduced with the kind permissions of the copyright owner. +<br/>Personal use of this material is permitted. However, permission to reprint/republish this +<br/>material for advertising or promotional purposes or for creating new collective works for +<br/>resale or redistribution to servers or lists, or to reuse any copyrighted component of this work +<br/>in other works must be obtained from the IEEE. +<br/>Copyright : 2008, IEEE +</td></tr><tr><td>22f656d0f8426c84a33a267977f511f127bfd7f3</td><td></td></tr><tr><td>2271d554787fdad561fafc6e9f742eea94d35518</td><td>TECHNISCHE UNIVERSIT ¨AT M ¨UNCHEN +<br/>Lehrstuhl f¨ur Mensch-Maschine-Kommunikation +<br/>Multimodale Mensch-Roboter-Interaktion +<br/>f¨ur Ambient Assisted Living +<br/>Tobias F. Rehrl +<br/>Vollst¨andiger Abdruck der von der Fakult¨at f¨ur Elektrotechnik und Informationstechnik +<br/>der Technischen Universit¨at M¨unchen zur Erlangung des akademischen Grades eines +<br/>Doktor-Ingenieurs (Dr.-Ing.) +<br/>genehmigten Dissertation. +<br/>Vorsitzende: +<br/>Pr¨ufer der Dissertation: 1. Univ.-Prof. Dr.-Ing. habil. Gerhard Rigoll +<br/>2. Univ.-Prof. Dr.-Ing. Horst-Michael Groß +<br/>Univ.-Prof. Dr.-Ing. Sandra Hirche +<br/>(Technische Universit¨at Ilmenau) +<br/>Die Dissertation wurde am 17. April 2013 bei der Technischen Universit¨at M¨unchen +<br/>eingereicht und durch die Fakult¨at f¨ur Elektrotechnik und Informationstechnik am +<br/>8. Oktober 2013 angenommen. +</td></tr><tr><td>22ec256400e53cee35f999244fb9ba6ba11c1d06</td><td></td></tr><tr><td>22a7f1aebdb57eecd64be2a1f03aef25f9b0e9a7</td><td></td></tr><tr><td>22e189a813529a8f43ad76b318207d9a4b6de71a</td><td>What will Happen Next? +<br/>Forecasting Player Moves in Sports Videos +<br/>UC Berkeley, STATS +<br/>UC Berkeley +<br/>UC Berkeley +</td></tr><tr><td>25d514d26ecbc147becf4117512523412e1f060b</td><td>Annotated Crowd Video Face Database +<br/>IIIT-Delhi, India +</td></tr><tr><td>25c19d8c85462b3b0926820ee5a92fc55b81c35a</td><td>Noname manuscript No. +<br/>(will be inserted by the editor) +<br/>Pose-Invariant Facial Expression Recognition +<br/>Using Variable-Intensity Templates +<br/>Received: date / Accepted: date +</td></tr><tr><td>258a8c6710a9b0c2dc3818333ec035730062b1a5</td><td>Benelearn 2005 +<br/>Annual Machine Learning Conference of +<br/>Belgium and the Netherlands +<br/>CTIT PROCEEDINGS OF THE FOURTEENTH +<br/>ANNUAL MACHINE LEARNING CONFERENCE +<br/>OF BELGIUM AND THE NETHERLANDS +</td></tr><tr><td>25695abfe51209798f3b68fb42cfad7a96356f1f</td><td>AN INVESTIGATION INTO COMBINING +<br/>BOTH FACIAL DETECTION AND +<br/>LANDMARK LOCALISATION INTO A +<br/>UNIFIED PROCEDURE USING GPU +<br/>COMPUTING +<br/> MSc by Research +<br/>2016 +</td></tr><tr><td>25d3e122fec578a14226dc7c007fb1f05ddf97f7</td><td>The First Facial Expression Recognition and Analysis Challenge +</td></tr><tr><td>2597b0dccdf3d89eaffd32e202570b1fbbedd1d6</td><td>Towards predicting the likeability of fashion images +</td></tr><tr><td>25982e2bef817ebde7be5bb80b22a9864b979fb0</td><td></td></tr><tr><td>25e05a1ea19d5baf5e642c2a43cca19c5cbb60f8</td><td>Label Distribution Learning +</td></tr><tr><td>2559b15f8d4a57694a0a33bdc4ac95c479a3c79a</td><td>570 +<br/>Contextual Object Localization With Multiple +<br/>Kernel Nearest Neighbor +<br/>Gert Lanckriet, Member, IEEE +</td></tr><tr><td>2574860616d7ffa653eb002bbaca53686bc71cdd</td><td></td></tr><tr><td>25f1f195c0efd84c221b62d1256a8625cb4b450c</td><td>1-4244-1017-7/07/$25.00 ©2007 IEEE +<br/>1091 +<br/>ICME 2007 +</td></tr><tr><td>25885e9292957feb89dcb4a30e77218ffe7b9868</td><td>JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2016 +<br/>Analyzing the Affect of a Group of People Using +<br/>Multi-modal Framework +</td></tr><tr><td>259706f1fd85e2e900e757d2656ca289363e74aa</td><td>Improving People Search Using Query Expansions +<br/>How Friends Help To Find People +<br/>LEAR - INRIA Rhˆone Alpes - Grenoble, France +</td></tr><tr><td>25728e08b0ee482ee6ced79c74d4735bb5478e29</td><td></td></tr><tr><td>258a2dad71cb47c71f408fa0611a4864532f5eba</td><td>Discriminative Optimization +<br/>of Local Features for Face Recognition +<br/> +<br/>H O S S E I N A Z I Z P O U R +<br/> +<br/>Master of Science Thesis +<br/>Stockholm, Sweden 2011 +<br/> +</td></tr><tr><td>25127c2d9f14d36f03d200a65de8446f6a0e3bd6</td><td>Journal of Theoretical and Applied Information Technology +<br/> 20th May 2016. Vol.87. No.2 +<br/>© 2005 - 2016 JATIT & LLS. All rights reserved. +<br/>ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195 +<br/>EVALUATING THE PERFORMANCE OF DEEP SUPERVISED +<br/>AUTO ENCODER IN SINGLE SAMPLE FACE RECOGNITION +<br/>PROBLEM USING KULLBACK-LEIBLER DIVERGENCE +<br/>SPARSITY REGULARIZER +<br/> Faculty of Computer of Computer Science, Universitas Indonesia, Kampus UI Depok, Indonesia +</td></tr></table></body></html>
\ No newline at end of file |