junshutang / EGGAN

[ICME 2020, Oral] Fine-Grained Expression Manipulation via Structured Latent Space

EGGAN: Fine-Grained Expression Manipulation via Structured Latent Space

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.

Dependencies

pip install -r requirement.txt

Dataset

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.

Pretrain process

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.

Training

bash runner.sh

Testing

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

bash test.sh

Citation

@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}
}