Pytorch implementation of multi-modality I2I translation for multi-domains. The project is an extension to the "Diverse Image-to-Image Translation via Disentangled Representations(https://arxiv.org/abs/1808.00948)", ECCV 2018. With the disentangled representation framework, we can learn diverse image-to-image translation among multiple domains. [DRIT]

Contact: Hsin-Ying Lee (hlee246@ucmerced.edu) and Hung-Yu Tseng (htseng6@ucmerced.edu)

• Python 3.5 or Python 3.6
• Pytorch 0.4.0 and torchvision (https://pytorch.org/)
• TensorboardX
• Tensorflow (for tensorboard usage)
• Docker file based on CUDA 9.0, CuDNN 7.1, and Ubuntu 16.04 is provided in the [DRIT] github page.

• Training

python train.py --dataroot DATAROOT --name NAME --num_domains NUM_DOMAINS --display_dir DISPLAY_DIR --result_dir RESULT_DIR --isDcontent

• Testing

python test.py --dataroot DATAROOT --name NAME --num_domains NUM_DOMAINS --out_dir OUT_DIR --resume MODEL_DIR --num NUM_PER_IMG

We validate our model on two datasets:

• art: Containing three domains: real images, Monet images, uki-yoe images. Data can be downloaded from CycleGAN website.
• weather: Containing four domains: sunny, cloudy, snowy, and foggy. Data is randomly selected from the Image2Weather dataset website.

The different domains in a dataset should be placed in folders "trainA, trainB, ..." in the alphabetical order.

• The pretrained model on the art dataset

bash ./models/download_model.sh art

• The pretrained model on the weather dataset

bash ./models/download_model.sh weather

• The feature transformation (i.e. concat 0) is not fully tested since both art and weather datasets do not require shape variations
• The hyper-parameters matter and are task-dependent. They are not carefully selected yet.
• Feel free to contact the author for any potential improvement of the code.

Diverse Image-to-Image Translation via Disentangled Representations
Hsin-Ying Lee*, Hung-Yu Tseng*, Jia-Bin Huang, Maneesh Kumar Singh, and Ming-Hsuan Yang
European Conference on Computer Vision (ECCV), 2018 (oral) (* equal contribution)

Please cite our paper if you find the code or dataset useful for your research.

@inproceedings{DRIT,
  author = {Lee, Hsin-Ying and Tseng, Hung-Yu and Huang, Jia-Bin and Singh, Maneesh Kumar and Yang, Ming-Hsuan},
  booktitle = {European Conference on Computer Vision},
  title = {Diverse Image-to-Image Translation via Disentangled Representations},
  year = {2018}
}