Neural_Artistic_Style

• Implementation of paper A Neural Algorithm of Artistic Style (arXiv:1508.06576) for re-construct an image with the art style of another image (Inspiration of codes drawn from anishatahlye's neural-style. For a better (also the first) implementation in Tensorflow, please refer to his repository)
• Implementation of paper Preserving Color in Neural Artistic Style Transfer (arXiv:1606.05897) for the option of preserving color scheme of the original image when tranferring styles

Dependecies

• Python 3+ (Might work with Python 2.7+, only tested in Python 3.5)
• Tensorflow
• Numpy
• Scipy
• A pretrained VGG19 convnet. The program uses imagenet pretrained vgg19. By default, the program assume the convnet lies within the same folder as paint_style.py

Example Usage

python3 paint_style.py --content CONTENT_IMAGE --styles STYLE_IMAGE(s) --output OUTPUT_IMAGE

or

python3 paint_style.py -c CONTENT_IMAGE -s STYLE_IMAGE(s) -o OUTPUT_IMAGE

Run with --preserve-colors True to preserve the original color schemes

More options exists. Run python3 paint_style.py --help for a decriptions of all available options.

Demo

The program was tested converting the following image into different artistic styles:

• The Starry Night, Vincent Van Gogh, 1989, Expressionism

• Impression, Sunrise, Claude Monet, 1872, Impressionism

• The Shipwreck of the Minotaur, Joseph Mallord William Turner, 1805, Romantism

• Bottle and Fishes, Georges Braque, 1912, Cubism

• Color Scheme Preserving

Preserve the original colr scheme while using the art style of "The Shipwreck of the Minotaur"