From 12bab78bc2100861d227b7ef0a1cb1e58ab11576 Mon Sep 17 00:00:00 2001 From: Cameron Date: Wed, 26 Oct 2016 14:12:35 -0600 Subject: Update README.md --- README.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/README.md b/README.md index e0a89fe..38ef912 100644 --- a/README.md +++ b/README.md @@ -209,14 +209,14 @@ Here we reproduce Figure 3 from [the original paper](https://arxiv.org/abs/1508. conv4_1 - + conv5_1 - + -- cgit v1.2.3-70-g09d2 From d5d1d2c36ccdc609b5a84dec63f9de74bb2ad602 Mon Sep 17 00:00:00 2001 From: Cameron Date: Thu, 27 Oct 2016 16:42:29 -0600 Subject: Update README.md --- README.md | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/README.md b/README.md index 38ef912..baf3250 100644 --- a/README.md +++ b/README.md @@ -432,3 +432,10 @@ If you find this code useful for your research, please cite: howpublished = {\url{https://github.com/cysmith/neural-style-tf}}, } ``` +

+
+If what AARON is making is not art, what is it exactly, and in what ways, other than its origin,
does it differ from the "real thing?" If it is not thinking, what exactly is it doing? +
 +- Harold Cohen (on his AI painter AARON), 1995 +

+ -- cgit v1.2.3-70-g09d2 From 0b361d97fae7c75f368464844850e79a880f3279 Mon Sep 17 00:00:00 2001 From: Cameron Date: Thu, 27 Oct 2016 18:17:53 -0600 Subject: Update README.md --- README.md | 7 ------- 1 file changed, 7 deletions(-) diff --git a/README.md b/README.md index baf3250..38ef912 100644 --- a/README.md +++ b/README.md @@ -432,10 +432,3 @@ If you find this code useful for your research, please cite: howpublished = {\url{https://github.com/cysmith/neural-style-tf}}, } ``` -

-
-If what AARON is making is not art, what is it exactly, and in what ways, other than its origin,
does it differ from the "real thing?" If it is not thinking, what exactly is it doing? -
 -- Harold Cohen (on his AI painter AARON), 1995 -

- -- cgit v1.2.3-70-g09d2 From 0a8a38c013e570524ce9ccaf68b67f8ffc5d242c Mon Sep 17 00:00:00 2001 From: Cameron Date: Thu, 27 Oct 2016 18:45:59 -0600 Subject: Update README.md --- README.md | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/README.md b/README.md index 38ef912..0ae6568 100644 --- a/README.md +++ b/README.md @@ -159,7 +159,7 @@ Animations can be rendered by applying the algorithm to each source frame. For

### Gradient Descent Initialization -The initialization of the gradient descent is controlled using `--init_img_type` for single images and `--init_frame_type` or `--first_frame_type` for video frames. For single images, the gradient descent can be initialized with white noise, the content image, or the style image. For video frames, the gradient descent can also be initialized with the previous stylized frame or the previous stylized frame warped to the current frame. White noise allows an arbitrary number of distinct images to be generated. Whereas, initializing with a fixed image always converges to the same output. +The initialization of the gradient descent is controlled using `--init_img_type` for single images and `--init_frame_type` or `--first_frame_type` for video frames. White noise allows an arbitrary number of distinct images to be generated. Whereas, initializing with a fixed image always converges to the same output. Here we reproduce Figure 6 from the first paper:

@@ -175,7 +175,7 @@ Here we reproduce Figure 6 from the first paper: *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). +The feature complexities and receptive field sizes increase down the VGG-199 CNN heirarchy. Here we reproduce Figure 3 from [the original paper](https://arxiv.org/abs/1508.06576): @@ -222,6 +222,8 @@ Here we reproduce Figure 3 from [the original paper](https://arxiv.org/abs/1508.
+*Rows*: increasing subsets of CNN layers +*Columns*: relative weighting of the the content and style reconstruction (see Content / Style Tradeoff) ## Setup #### Dependencies: -- cgit v1.2.3-70-g09d2 From 94cb0cf771a9f24c19b198d1185ae98733f363ef Mon Sep 17 00:00:00 2001 From: Cameron Date: Thu, 27 Oct 2016 18:46:35 -0600 Subject: Update README.md --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index 0ae6568..d2f157f 100644 --- a/README.md +++ b/README.md @@ -222,7 +222,7 @@ Here we reproduce Figure 3 from [the original paper](https://arxiv.org/abs/1508. -*Rows*: increasing subsets of CNN layers +*Rows*: increasing subsets of CNN layers *Columns*: relative weighting of the the content and style reconstruction (see Content / Style Tradeoff) ## Setup -- cgit v1.2.3-70-g09d2 From 3d83f5a8f2f2ca11f861c86d5272f021a5c042cb Mon Sep 17 00:00:00 2001 From: Cameron Date: Thu, 27 Oct 2016 20:44:59 -0600 Subject: Update README.md --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index d2f157f..3eeaffa 100644 --- a/README.md +++ b/README.md @@ -175,7 +175,7 @@ Here we reproduce Figure 6 from the first paper: *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 feature complexities and receptive field sizes increase down the CNN heirarchy. Here we reproduce Figure 3 from [the original paper](https://arxiv.org/abs/1508.06576): -- cgit v1.2.3-70-g09d2 From 82c9f60651d9633f0a18f74775b818d4418cbece Mon Sep 17 00:00:00 2001 From: Cameron Date: Thu, 27 Oct 2016 23:17:35 -0600 Subject: Update README.md --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index 3eeaffa..00151ac 100644 --- a/README.md +++ b/README.md @@ -17,7 +17,7 @@ The Neural Style algorithm combines the content of one image with the style of a

-Transfering the style of various artworks to the same content image produces qualitatively convincing results: +Transferring the style of various artworks to the same content image produces qualitatively convincing results:

-- cgit v1.2.3-70-g09d2 From 22cd8c6ec8fd7ac741fed51e682bec5f569bba19 Mon Sep 17 00:00:00 2001 From: Cameron Date: Fri, 28 Oct 2016 09:23:21 -0600 Subject: Update README.md --- README.md | 52 ++++++++++++++++++++++++++-------------------------- 1 file changed, 26 insertions(+), 26 deletions(-) diff --git a/README.md b/README.md index 00151ac..681bb0f 100644 --- a/README.md +++ b/README.md @@ -115,37 +115,37 @@ The algorithm is not constrained to artistic painting styles. It can also be ap ### Segmentation Style can be transferred to semantic segmentations in the content image.

- - - - - - - - - - - - - + + + + + + + + + + + + +

Multiple styles can be transferred to the foreground and background of the content image.

- - - - - - - - - - - - - + + + + + + + + + + + + +

*Left to right*: content image, foreground style, background style, foreground mask, background mask, stylized image -- cgit v1.2.3-70-g09d2 From a44d1c3e9631bc20f84823f9495b5f43583e94b6 Mon Sep 17 00:00:00 2001 From: Cameron Date: Sat, 29 Oct 2016 17:43:00 -0600 Subject: Update README.md --- README.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/README.md b/README.md index 681bb0f..75cbdea 100644 --- a/README.md +++ b/README.md @@ -222,8 +222,8 @@ Here we reproduce Figure 3 from [the original paper](https://arxiv.org/abs/1508.
-*Rows*: increasing subsets of CNN layers -*Columns*: relative weighting of the the content and style reconstruction (see Content / Style Tradeoff) +*Rows*: increasing subsets of CNN layers; i.e. 'conv4_1' means using 'conv1_1', 'conv2_1', 'conv3_1', 'conv4_1'. +*Columns*: alpha/beta ratio of the the content and style reconstruction (see Content / Style Tradeoff) ## Setup #### Dependencies: -- cgit v1.2.3-70-g09d2 From 70782471fc3f8fa12971b4ea4c75a254c2f58a4e Mon Sep 17 00:00:00 2001 From: Cameron Date: Sat, 29 Oct 2016 17:43:43 -0600 Subject: Update README.md --- README.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/README.md b/README.md index 75cbdea..4f082c3 100644 --- a/README.md +++ b/README.md @@ -222,8 +222,8 @@ Here we reproduce Figure 3 from [the original paper](https://arxiv.org/abs/1508. -*Rows*: increasing subsets of CNN layers; i.e. 'conv4_1' means using 'conv1_1', 'conv2_1', 'conv3_1', 'conv4_1'. -*Columns*: alpha/beta ratio of the the content and style reconstruction (see Content / Style Tradeoff) +*Rows*: increasing subsets of CNN layers; i.e. 'conv4_1' means using 'conv1_1', 'conv2_1', 'conv3_1', 'conv4_1'. +*Columns*: alpha/beta ratio of the the content and style reconstruction (see Content / Style Tradeoff). ## Setup #### Dependencies: -- cgit v1.2.3-70-g09d2 From d03c67a4943ba21cb85450b5034b7bd0e7589094 Mon Sep 17 00:00:00 2001 From: Cameron Date: Tue, 1 Nov 2016 23:33:10 -0600 Subject: Add files via upload --- examples/equations/email.png | Bin 0 -> 425 bytes 1 file changed, 0 insertions(+), 0 deletions(-) create mode 100644 examples/equations/email.png diff --git a/examples/equations/email.png b/examples/equations/email.png new file mode 100644 index 0000000..8b8d0fc Binary files /dev/null and b/examples/equations/email.png differ -- cgit v1.2.3-70-g09d2 From 86693f0abc1a4420af6470b8ebce35ab3fd308d9 Mon Sep 17 00:00:00 2001 From: Cameron Date: Tue, 1 Nov 2016 23:38:30 -0600 Subject: Update README.md --- README.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/README.md b/README.md index 4f082c3..0f5f9c2 100644 --- a/README.md +++ b/README.md @@ -338,7 +338,7 @@ python neural_style.py --video \ * `--img_name`: Filename of the output image. *Default*: `result` * `--verbose`: Boolean flag indicating if statements should be printed to the console. -#### Optimization Arguments +"#### Optimization Arguments * `--optimizer`: Loss minimization optimizer. L-BFGS gives better results. Adam uses less memory. *Choices*: `lbfgs`, `adam`. *Default*: `lbfgs` * `--learning_rate`: Learning-rate parameter for the Adam optimizer. *Default*: `1e1` * `--max_iterations`: Max number of iterations for the Adam or L-BFGS optimizer. *Default*: `1000` @@ -366,7 +366,7 @@ python neural_style.py --video \ ## Questions and Errata -Send questions or issues: cysmith1010@gmail.com +Send questions or issues: ## Memory By default, `neural-style-tf` uses the NVIDIA cuDNN GPU backend for convolutions and L-BFGS for optimization. -- cgit v1.2.3-70-g09d2 From 8b5e3b0d37c66844d2ba0b50348070f45688c4d9 Mon Sep 17 00:00:00 2001 From: Cameron Date: Wed, 2 Nov 2016 00:35:07 -0600 Subject: Update README.md --- README.md | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/README.md b/README.md index 0f5f9c2..dbdece9 100644 --- a/README.md +++ b/README.md @@ -131,7 +131,6 @@ Style can be transferred to semantic segmentations in the content image.

Multiple styles can be transferred to the foreground and background of the content image. -

@@ -338,7 +337,7 @@ python neural_style.py --video \ * `--img_name`: Filename of the output image. *Default*: `result` * `--verbose`: Boolean flag indicating if statements should be printed to the console. -"#### Optimization Arguments +#### Optimization Arguments * `--optimizer`: Loss minimization optimizer. L-BFGS gives better results. Adam uses less memory. *Choices*: `lbfgs`, `adam`. *Default*: `lbfgs` * `--learning_rate`: Learning-rate parameter for the Adam optimizer. *Default*: `1e1` * `--max_iterations`: Max number of iterations for the Adam or L-BFGS optimizer. *Default*: `1000` @@ -366,7 +365,8 @@ python neural_style.py --video \ ## Questions and Errata -Send questions or issues: +Send questions or issues: + ## Memory By default, `neural-style-tf` uses the NVIDIA cuDNN GPU backend for convolutions and L-BFGS for optimization. -- cgit v1.2.3-70-g09d2