Implementation of the multi-label version of Weakly-Supervised CNN Learning for Facial Action Unit Intensity Estimation, CVPR 2018.
- python
- tensorflow
- Collect data
- images: images are collected from the videos.
- tuples: detect the intensity peaks and valleys first and then sample tuples from each segment.
- Training
python ShollowCNN_AUModel_weak_train.py
- Testing
python ShollowCNN_AUModel_eval_weak.py
If the paper inspires you, please cite our work.
@inproceedings{zhang2018weakly,
title={Weakly-supervised deep convolutional neural network learning for facial action unit intensity estimation},
author={Zhang, Yong and Dong, Weiming and Hu, Bao-Gang and Ji, Qiang},
booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
pages={2314--2323},
year={2018}
}