An implementation of 'A Neural Algorithm of Artistic Style'
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
LICENSE
README.md
train_gogh.py
vgg16.model

README.md

An implementation of 'A Neural Algorithm of Artistic Style' (http://arxiv.org/abs/1508.06576)

This is an implementation of 'A Neural Algorithm of Artistic Style' (http://arxiv.org/abs/1508.06576). The code is largely based on https://github.com/pfnet-research/chainer-gogh and https://github.com/yusuketomoto/chainer-fast-neuralstyle

The script generates a stylised image from an input image and a style image.

Demo on Google Colaboratory

You can try the demo on a browser Google Colaboratory

Requirements

  • Python 3: Anaconda is recommended
  • Python libraries: Chainer, cupy: pip install cupy chainer
  • CUDA supported GPU is highly recommended. Without one, it takes ages to do anything with CNN.

How to use

python train.py -h

gives a brief description of command line arguments.

A typical conversion is performed by

python train_gogh.py input.jpg -s style.jpg -rs

Generated images will be found under the directory named 'result'.