August 17, 2019

137 words 1 min read

ceshine/fast-neural-style

ceshine/fast-neural-style

Pytorch Implementation of Perceptual Losses for Real-Time Style Transfer and Super-Resolution

repo name ceshine/fast-neural-style
repo link https://github.com/ceshine/fast-neural-style
homepage
language Jupyter Notebook
size (curr.) 12705 kB
stars (curr.) 134
created 2017-07-31
license

fast-neural-style

July 2018 Update:

  1. Upgrade to PyTorch 0.4.0.
  2. Minor Refactor.
  3. Allow assigning an independent style weight for each layer.
  4. Provide a Dockerfile

The code to the first blog post about this project can be found in tag 201707.

Stylize Script Usage Example

python stylize.py models/model_rain_princess_cropped.pth content_images/pic.jpg pic-512.jpg --resize=512

Old README

This personal fun project is heavily based on abhiskk/fast-neural-style with some changes and a video generation notebook/script:

  1. Use the official pre-trained VGG model
  2. Output intermediate results during training
  3. Add Total Variation Regularization as described in the paper

The model uses the method described in Perceptual Losses for Real-Time Style Transfer and Super-Resolution along with Instance Normalization.

comments powered by Disqus