------------ status: published title: Duke Multi-Target, Multi-Camera Tracking desc: Duke MTMC is a dataset of surveillance camera footage of students on Duke University campus subdesc: Duke MTMC contains over 2 million video frames and 2,000 unique identities collected from 8 HD cameras at Duke University campus in March 2014 slug: duke_mtmc cssclass: dataset image: assets/background.jpg published: 2019-2-23 updated: 2019-2-23 authors: Adam Harvey ------------ ### sidebar + Created: 2014 + Identities: Over 2,700 + Used for: Face recognition, person re-identification + Created by: Computer Science Department, Duke University, Durham, US + Website: duke.edu ## Duke Multi-Target, Multi-Camera Tracking Dataset (Duke MTMC) [ PAGE UNDER DEVELOPMENT ] Duke MTMC is a dataset of video recorded on Duke University campus during for the purpose of training, evaluating, and improving *multi-target multi-camera tracking*. The videos were recorded during February and March 2014 and cinclude Includes a total of 888.8 minutes of video (ind. verified) "We make available a new data set that has more than 2 million frames and more than 2,700 identities. It consists of 8×85 minutes of 1080p video recorded at 60 frames per second from 8 static cameras deployed on the Duke University campus during periods between lectures, when pedestrian traffic is heavy." The dataset includes approximately 2,000 annotated identities appearing in 85 hours of video from 8 cameras located throughout Duke University's campus. ![caption: Duke MTMC pixel-averaged image of camera #5 is shown with the bounding boxes for each student drawn in white. (c) Adam Harvey](assets/duke_mtmc_cam5_average_comp.jpg) According to the dataset authors, {% include 'map.html' %} {% include 'chart.html' %} {% include 'piechart.html' %} {% include 'supplementary_header.html' %} {% include 'citations.html' %} ---- ## Research Notes - "We make available a new data set that has more than 2 million frames and more than 2,700 identities. It consists of 8×85 minutes of 1080p video recorded at 60 frames per second from 8 static cameras deployed on the Duke University campus during periods between lectures, when pedestrian traffic is heavy." - 27a2fad58dd8727e280f97036e0d2bc55ef5424c - "This work was supported in part by the EPSRC Programme Grant (FACER2VM) EP/N007743/1, EPSRC/dstl/MURI project EP/R018456/1, the National Natural Science Foundation of China (61373055, 61672265, 61602390, 61532009, 61571313), Chinese Ministry of Education (Z2015101), Science and Technology Department of Sichuan Province (2017RZ0009 and 2017FZ0029), Education Department of Sichuan Province (15ZB0130), the Open Research Fund from Province Key Laboratory of Xihua University (szjj2015-056) and the NVIDIA GPU Grant Program." - ec9c20ed6cce15e9b63ac96bb5a6d55e69661e0b - "DukeMTMC aims to accelerate advances in multi-target multi-camera tracking. It provides a tracking system that works within and across cameras, a new large scale HD video data set recorded by 8 synchronized cameras with more than 7,000 single camera trajectories and over 2,000 unique identities, and a new performance evaluation method that measures how often a system is correct about who is where" - DukeMTMC is a new, manually annotated, calibrated, multi-camera data set recorded outdoors on the Duke University campus with 8 synchronized cameras. It consists of: 8 static cameras x 85 minutes of 1080p 60 fps video More than 2,000,000 manually annotated frames More than 2,000 identities Manual annotation by 5 people over 1 year More identities than all existing MTMC datasets combined Unconstrained paths, diverse appearance - DukeMTMC Project Ergys Ristani Ergys Ristani Ergys Ristani Ergys Ristani Ergys Ristani People involved: Ergys Ristani, Francesco Solera, Roger S. Zou, Rita Cucchiara, Carlo Tomasi. Navigation: Data Set Downloads Downloads Dataset Extensions Performance Measures Tracking Systems Publications How to Cite Contact Welcome to the Duke Multi-Target, Multi-Camera Tracking Project. DukeMTMC aims to accelerate advances in multi-target multi-camera tracking. It provides a tracking system that works within and across cameras, a new large scale HD video data set recorded by 8 synchronized cameras with more than 7,000 single camera trajectories and over 2,000 unique identities, and a new performance evaluation method that measures how often a system is correct about who is where. DukeMTMC Data Set Snapshot from the DukeMTMC data set. DukeMTMC is a new, manually annotated, calibrated, multi-camera data set recorded outdoors on the Duke University campus with 8 synchronized cameras. It consists of: 8 static cameras x 85 minutes of 1080p 60 fps video More than 2,000,000 manually annotated frames More than 2,000 identities Manual annotation by 5 people over 1 year More identities than all existing MTMC datasets combined Unconstrained paths, diverse appearance News 05 Feb 2019 We are organizing the 2nd Workshop on MTMCT and ReID at CVPR 2019 25 Jul 2018: The code for DeepCC is available on github 28 Feb 2018: OpenPose detections now available for download 19 Feb 2018: Our DeepCC tracker has been accepted to CVPR 2018 04 Oct 2017: A new blog post describes ID measures of performance 26 Jul 2017: Slides from the BMTT 2017 workshop are now available 09 Dec 2016: DukeMTMC is now hosted on MOTChallenge DukeMTMC Downloads DukeMTMC dataset (tracking) Dataset Extensions Below is a list of dataset extensions provided by the community: DukeMTMC-VideoReID (download) DukeMTMC-reID (download) DukeMTMC4REID DukeMTMC-attribute If you use or extend DukeMTMC, please refer to the license terms. DukeMTMCT Benchmark DukeMTMCT is a tracking benchmark hosted on motchallenge.net. Click here for the up-to-date rankings. Here you will find the official motchallenge-devkit used for evaluation by MOTChallenge. For detailed instructions how to submit on motchallenge you can refer to this link. Trackers are ranked using our identity-based measures which compute how often the system is correct about who is where, regardless of how often a target is lost and reacquired. Our measures are useful in applications such as security, surveillance or sports. This short post describes our measures with illustrations, while for details you can refer to the original paper. Tracking Systems We provide code for the following tracking systems which are all based on Correlation Clustering optimization: DeepCC for single- and multi-camera tracking [1] Single-Camera Tracker (demo video) [2] Multi-Camera Tracker (demo video, failure cases) [2] People-Groups Tracker [3] Original Single-Camera Tracker [4]