Home

Awesome

Paper

Yihao Zhao, Ruihai Wu, Hao Dong, "Unpaired Image-to-Image Translation using Adversarial Consistency Loss", ECCV 2020

arXiv: https://arxiv.org/abs/2003.04858

Code usage

For environment:

conda env create -f acl-gan.yaml

For training:

python train.py --config configs/male2female.yaml

For test:

python test.py --config configs/male2female.yaml --input inputs/test_male.jpg --checkpoint ./models/test.pth

Experimental Results

Ablation study
<img src="figures/ablation_study.png" alt="ablation_study" style="zoom:50%;" />
Male-to-female
<img src="figures/male2female.png" alt="male2female" style="zoom:50%;" />
Glasses Removal
<img src="figures/glasses_removal.png" alt="glasses_removal" style="zoom:50%;" />
Selfie-to-anime
<img src="figures/selfie2anime.png" alt="selfie2anime" style="zoom:50%;" />

For more results, please refer to our paper.

Citation

If you find this code useful for your research, please cite our paper:

@inproceedings{zhao2020aclgan,
  title={Unpaired Image-to-Image Translation using Adversarial Consistency Loss},
  author={Zhao, Yihao and Wu, Ruihai and Dong, Hao},
  booktitle={ECCV},
  year={2020}
}