/Former-DFER

“Former-DFER: Dynamic Facial Expression Recognition Transformer”, ACM MM 2021.

Primary LanguagePythonMIT LicenseMIT

Former-DFER

Zengqun Zhao, Qingshan Liu. "Former-DFER: Dynamic Facial Expression Recognition Transformer". ACM International Conference on Multimedia.

Requirements

  • pytorch==1.8.0
  • torchvision==0.9.0

Training

  • Step 1: download DFEW dataset.
  • Step 2: fill in the your_DFEW_Face_path in script.py, then run script.py.
  • Step 3: run sh DFEW_Five_Fold.sh

Pre-trained Models

The pre-trained Former-DFER model on DFEW can be downloaded here.

Citation

@inproceedings{zhao2021former,
  title={Former-DFER: Dynamic Facial Expression Recognition Transformer},
  author={Zhao, Zengqun and Liu, Qingshan},
  booktitle={Proceedings of the 29th ACM International Conference on Multimedia},
  pages={1553--1561},
  year={2021}
}