Everyone_Is_Van_Gogh

Everyone_Is_Van_Gogh is a tensorflow implementation of style transfer which described in the next paper:

• A Neural Algorithm of Artistic Style

And I use VGG19 which was proposed in this paper:

• Very Deep Convolutional Networks for Large-Scale Image Recognition

To use the VGG19 networks, you have to download the npy files for VGG19 NPY from here.

Environment

• 1.tensorflow v1.4
• 2.Cudnn v6.0
• 3.Python v3.6
• 4.Python packages:numpy,PIL,scipy
• 4.Bash(gitssh)

Getting Started

• 1.Clone this project : git clone [email protected]:hualin95/Everyone_Is_Van_Gogh.git
• 2.Enter the folder Everyone_Is_Van_Gogh : cd Everyone_Is_Van_Gogh
• 3.Run the file train.py :

python train.py --VGG19_Model_Path [your filepath of vgg19.npy]  --content_image [your filepath of content_image] --style_image [your filepath of style_image]

For example (I have move the file vgg19.npy to the folder Everyone_Is_Van_Gogh):

python train.py --VGG19_Model_Path "vgg19.npy"  --content_image "images/content/johannesburg.jpg" --style_image "images/style/starry-night.jpg"

More : I guess you want to train the project with GPU if you have one, so you can run the file train.py like this:

CUDA_VISIBLE_DEVICES=[GPU_num] python train.py --VGG19_Model_Path [your filepath of vgg19.npy]  --content_image [your filepath of content_image] --style_image [your filepath of style_image]

For example :

CUDA_VISIBLE_DEVICES=4 python train.py --VGG19_Model_Path "vgg19.npy"  --content_image "images/content/johannesburg.jpg" --style_image "images/style/starry-night.jpg"

• 4.The generated image is saved in Everyone_Is_Van_Gogh/images/generated/

Sample

Style_iamge

The Starry Night is an oil on canvas by the Dutch post-impressionist painter Vincent van Gogh in June 1889.

Content_image

This is picture of University of Chinese Academy of Sciences where I am studying now and finish this project.

Generated_image(after 1000 iteration)