CycleGAN

New: We also add a simple PyTorch implementation, see the "pytorch" folder

Tensorflow implementation of CycleGAN, mostly modified from https://github.com/XHUJOY/CycleGAN-tensorflow to a simpler version

Paper: Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networkssee Author: Jun-Yan Zhu et al.

Exemplar results on testset

• gif: horse -> zebra
• row 1: horse -> zebra -> reconstructed horse, row 2: zebra -> horse -> reconstructed zebra
• row 1: apple -> orange -> reconstructed apple, row 2: orange -> apple -> reconstructed orange

Prerequisites

• tensorflow r1.0
• python 2.7

Usage

cd CycleGAN-Tensorflow-PyTorch-Simple-master

Download Datasets

• Download the horse2zebra dataset:

sh ./download_dataset.sh horse2zebra

• Download the apple2orange dataset:

sh ./download_dataset.sh apple2orange

• See download_dataset.sh for more datasets

Train Example

python train.py --dataset=horse2zebra --gpu_id=0

Test Example

python test.py --dataset=horse2zebra

Gif2Gif

python gif2gif.py --gif=./pics/horse.gif --save_path=./pics/horse2zebra.gif --dataset=horse2zebra --direction=a2b