Pytorch implementation of CartoonGAN [1] (CVPR 2018)
- Parameters without information in the paper were set arbitrarily.
- I used face-cropped celebA (src) and anime (tgt) collected from the web data because I could not find the author's data.
python CartoonGAN.py --name your_project_name --src_data src_data_path --tgt_data tgt_data_path --vgg_model pre_trained_VGG19_model_path
The following shows basic folder structure.
├── data
│ ├── src_data # src data (not included in this repo)
│ │ ├── train
│ │ └── test
│ └── tgt_data # tgt data (not included in this repo)
│ ├── train
│ └── pair # edge-promoting results to be saved here
│
├── CartoonGAN.py # training code
├── edge_promoting.py
├── utils.py
├── networks.py
└── name_results # results to be saved here
| Input - Result (this repo) |
|
|
|
- I got the author's results from CaroonGAN-Test-Pytorch-Torch.
| Input - Result (this repo) | Author's pre-trained model (Hayao) | Author's pre-trained model (Hosoda) |
| 
| 
|
| 
| 
|
| 
| 
|
| 
| 
|
| 
| 
|
- NVIDIA GTX 1080 ti
- cuda 8.0
- python 3.5.3
- pytorch 0.4.0
- torchvision 0.2.1
- opencv 3.2.0
[1] Chen, Yang, Yu-Kun Lai, and Yong-Jin Liu. "CartoonGAN: Generative Adversarial Networks for Photo Cartoonization." Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2018.
(Full paper: http://openaccess.thecvf.com/content_cvpr_2018/papers/Chen_CartoonGAN_Generative_Adversarial_CVPR_2018_paper.pdf)