syp1997 / TransGAN

[Preprint] "TransGAN: Two Transformers Can Make One Strong GAN", Yifan Jiang, Shiyu Chang, Zhangyang Wang

TransGAN: Two Transformers Can Make One Strong GAN

Code used for TransGAN: Two Transformers Can Make One Strong GAN.

News

checkpoint for generating images on celeba64 dataset is released!

Main Pipeline

Visual Results

prepare fid statistic file

mkdir fid_stat

Download the pre-calculated statistics (Google Drive) to ./fid_stat.

Environment

pip install -r requirements.txt

Notice: Pytorch version has to be <=1.3.0 !

Training

Coming soon

Testing

Firstly download the checkpoint from (Google Drive) to ./pretrained_weight

# cifar-10
sh exps/cifar10_test.sh

# stl-10
sh exps/stl10_test.sh

# celeba64
sh exps/celeba64_test.sh

Acknowledgement

Codebase from AutoGAN, pytorch-image-models

Citation

if you find this repo is helpful, please cite

@article{jiang2021transgan,
  title={TransGAN: Two Transformers Can Make One Strong GAN},
  author={Jiang, Yifan and Chang, Shiyu and Wang, Zhangyang},
  journal={arXiv preprint arXiv:2102.07074},
  year={2021}
}