This repository implements the training and testing of EGGAN for "Fine-Grained Expression Manipulation via Structured Latent Space". It offers the original implementation of the paper in PyTorch.

pip install -r requirement.txt

The data of DISFA dataset can be download from here. Please unzip the data.zip under the "data/DISFA" folder, in which the path file of DISFA contains each frame of videos.

The pretrained model of the Identity Classifier can be download from here. You can train with other datasets by:

python3 pretrain.py

And modify the trainning details.

bash runner.sh

After trainning, you can test the model by loading the specified model.

bash test.sh

@inproceedings{tang2020fine,
  title={Fine-Grained Expression Manipulation Via Structured Latent Space},
  author={Tang, Junshu and Shao, Zhiwen and Ma, Lizhuang},
  booktitle={2020 IEEE International Conference on Multimedia and Expo (ICME)},
  pages={1--6},
  year={2020},
  organization={IEEE}
}