From bb968cab6afee2a80aa841588477a410fb6d66d4 Mon Sep 17 00:00:00 2001 From: Cameron Date: Wed, 26 Oct 2016 00:21:53 -0600 Subject: Update README.md --- README.md | 50 +++++++++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 49 insertions(+), 1 deletion(-) diff --git a/README.md b/README.md index ef15203..2cbd54c 100644 --- a/README.md +++ b/README.md @@ -42,7 +42,6 @@ Here we reproduce Figure 3 from the first paper, which renders a photograph of t

### Content / Style Tradeoff - The relative weight of the style and content can be controlled. Here we render with an increasing style weight applied to [Red Canna](http://www.georgiaokeeffe.net/red-canna.jsp): @@ -175,6 +174,55 @@ Here we reproduce Figure 6 from the first paper: *Top row (left to right)*: Initialized with the content image, the style image, white noise (RNG seed 1) *Bottom row (left to right)*: Initialized with white noise (RNG seed 2), white noise (RNG seed 3), white noise (RNG seed 4) +### Layer Representations +The feature complexities and receptive field sizes increase down the VGG-199 CNN heirarchy. The rows in the below figure show the increasing complexity and size of local image structures as an increasing subset of CNN layers are used. The columns show the alpha/beta ratio which is the relative weighting of the the content and style reconstruction (see Content / Style Tradeoff). + +Here we reproduce Figure 3 from [the original paper](https://arxiv.org/abs/1508.06576): + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +
1 x 10^-51 x 10^-41 x 10^-31 x 10^-2
relu1_1
relu2_1
relu3_1
relu4_1
relu5_1
+ ## Setup #### Dependencies: * [tensorflow](https://github.com/tensorflow/tensorflow) -- cgit v1.2.3-70-g09d2 From 5331725909db9fbc496380eba78492301051755e Mon Sep 17 00:00:00 2001 From: Cameron Date: Wed, 26 Oct 2016 01:54:44 -0600 Subject: Update README.md --- README.md | 42 +++++++++++++++++++++--------------------- 1 file changed, 21 insertions(+), 21 deletions(-) diff --git a/README.md b/README.md index 2cbd54c..706f8cd 100644 --- a/README.md +++ b/README.md @@ -172,7 +172,7 @@ Here we reproduce Figure 6 from the first paper:

*Top row (left to right)*: Initialized with the content image, the style image, white noise (RNG seed 1) -*Bottom row (left to right)*: Initialized with white noise (RNG seed 2), white noise (RNG seed 3), white noise (RNG seed 4) +*Bottom row (left to right)*: Initialized with white noise (RNG seeds 2, 3, 4) ### Layer Representations The feature complexities and receptive field sizes increase down the VGG-199 CNN heirarchy. The rows in the below figure show the increasing complexity and size of local image structures as an increasing subset of CNN layers are used. The columns show the alpha/beta ratio which is the relative weighting of the the content and style reconstruction (see Content / Style Tradeoff). @@ -187,39 +187,39 @@ Here we reproduce Figure 3 from [the original paper](https://arxiv.org/abs/1508. 1 x 10^-2 -relu1_1 +conv1_1 - - - + + + -relu2_1 +conv2_1 - - - + + + -relu3_1 +conv3_1 - - - + + + -relu4_1 +conv4_1 - - - + + + -relu5_1 +conv5_1 - - - + + + -- cgit v1.2.3-70-g09d2