chainer neural-style & fast-neural-style

Chainer implementation of A Neural Algorithm of Artistic Style and Perceptual Losses for Real-Time Style Transfer and Super-Resolution.

neural style

Download VGG-19

Download the original VGG-19 model and then convert it to Chainer model:

python caffe_model_2_pickle.py

Or Download the converted VGG-19 Chainer model here.

Now you're an artist!!

 python neural_style.py --content_image [Content Image] --style_images [Style Image;Style Image;...] 

Note: python neural_style -h for more details.

Your Gallery

• Content Image

  

• Style Image

  

• Result

  • --original_color False --style_color False

    

  • --original_color True

    

  • --style_color True

    

• Content Image

  

• Style Images

  

  

• Result Image

  

fast neural style

Download Dataset

Download COCO training set and validation set. unzip training set and validation set under the same root folder.

Download VGG-19

See above.

Training using Batch Normalization

python fast_neural_style.py --data_root [COCO ROOT FOLDER] --instance_normalization False

• Training Curve

Training using Instance Normalization

python fast_neural_style.py --data_root [COCO ROOT FOLDER] --instance_normalization True

• Training Curve

Gallery

• Batch Normalization

  

  

  

• Instance Normalization