A PyTorch implementation of neural style transfer.
Check out the paper here.
To run the code quickly:
pip3 install -r requirements.txt
# run with randomly select content and style images
python3 neural.py
if __name__ == '__main__':
# generate images, create the neural style object
neural_style_system = NeuralStyle()
# get randon style+content image
# neural_style_system.get_img(content_img_name='ma3.jpg',
# style_img_name='flowercarrier.jpg')
neural_style_system.get_img()
neural_style_system.plot_content_then_style()
# select the model (by default, we select the vgg-19 model)
neural_style_system.select_model(model_selection='vgg19')
# style weight is how much to prioritize style over content
# higher style weight = focus more on matching the style picture
neural_style_system.run_style_transfer(style_weight=1000, content_weight=5)
Better documentation coming soon!