https://github.com/wkentaro/chainer-bicyclegan

Chainer implementation of "Toward Multimodal Image-to-Image Translation".
https://github.com/wkentaro/chainer-bicyclegan

chainer computer-graphics computer-vision deep-learning gan generative-adversarial-network image-generation pix2pix

Last synced: 4 months ago
JSON representation

Chainer implementation of "Toward Multimodal Image-to-Image Translation".

• Host: GitHub
• URL: https://github.com/wkentaro/chainer-bicyclegan
• Owner: wkentaro
• Created: 2017-12-28T16:54:51.000Z (about 7 years ago)
• Default Branch: master
• Last Pushed: 2018-03-22T09:32:31.000Z (almost 7 years ago)
• Last Synced: 2024-10-05T06:08:01.582Z (4 months ago)
• Topics: chainer, computer-graphics, computer-vision, deep-learning, gan, generative-adversarial-network, image-generation, pix2pix
• Language: Python
• Homepage:
• Size: 2.84 MB
• Stars: 7
• Watchers: 4
• Forks: 2
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# chainer-bicyclegan

[![Build Status](https://travis-ci.org/wkentaro/chainer-bicyclegan.svg?branch=master)](https://travis-ci.org/wkentaro/chainer-bicyclegan)

Chainer implementation of ["Toward Multimodal Image-to-Image Translation"](https://arxiv.org/abs/1711.11586).  

This is a faithful re-implementation of [the official PyTorch implementation](https://github.com/junyanz/BicycleGAN).

## Installation

```bash

git clone --recursive https://github.com/wkentaro/chainer-bicyclegan.git

cd chainer-bicyclegan

conda install -c menpo -y opencv

pip install .

```

## Training

```bash

cd examples/bicyclegan

# ./train.py  --gpu 

./train.py edges2shoes --gpu 0

```

## Results

It shows that the BicycleGAN output has more variety with smaller batchsize.

### edges2shoes

- batchsize=2,epoch=60 (left)

- batchsize=60,epoch=60 (right)

 

### edges2handbags

- batchsize=2,epoch=5 (left)

- batchsize=60,epoch=15 (right)