StarGAN
Tensorflow implementation of StarGAN: Unified Generative Adversarial Networks for Multi-Domain Image-to-Image Translation
Prerequisites
- Python 3.6
- Tensorflow 1.6
Results for Facial Attribute Transfer on CelebA
The images are generated by StarGAN trained on the CelebA dataset.
Usage
1. Clone the repository
$ git clone https://github.com/ioanachelu/StarGAN.git
$ cd StarGAN/
./data
. It should have the following substructure:
2. Download the dataset into .data
celebA
list_attr_celebs.txt
3. Train StarGAN with CelebA
$ python train.py --train=True --logdir=./logdir --config=default --dataset_path=../data
4. Resume StarGAN training with CelebA
$ python train.py --train=True --resume=True --load_from=./logdir/0-default --config=default --dataset_path=../data
4. Test StarGAN
$ python run_wild.py --logdir=./logdir --config=default --load_from=./logdir/0-default --dataset_type=test