Facial Prior Guided Micro-Expression Generation

Implementation of FOMM with AWP and MRAA with AWP.

0. Table of Contents

• 0. Table of Contents

• 1. Authors & Maintainers

• 2. Change Log

• 3. Visualizations

• 4. Run the Code

• 5. License

1. Authors & Maintainers

• Yi Zhang|@zylye123
• Xinhua Xu|@sysu19351158
• Youjun Zhao|@zhaoyjoy
• Yuhang Wen|@Necolizer
• Zixuan Tang|@sysu19351118

2. Change Log

• [2022/02/18] Upload code.
• [2023/05/02] + Case Visualizations.

3. Visualizations

CASE	Driving	FOMM	FOMM w/ EWP	FOMM w/ AWP	MRAA	MRAA w/ AWP
1
2
3
4
5

• To facilitate micro-expression observations and reduce the storage space of this repository, all Gifs have been slowed down and compressed.
• Case 1-3: Positive. Case 4: Negative. Case 5: Surprise.

4. Run the Code

1. Operations of FOMM_with_AWP and MRAA_with_AWP are the same.

2. Prepare your dataset. Recommend CASME2, SAMM, SMIC-HS

   Divide into your_dataset/train and your_dataset/test

   Create or modify yaml format file your_dataset_train.yaml in ./config

3. Train

   python run.py --config config/your_dataset_train.yaml

   Log, parameters and checkpoints would be saved in ./log

4. Test

   Create or modify csv format file your_dataset_test.csv in ./data

   python run.py --config config/my_dataset_test.yaml --mode animate --checkpoint path/to/checkpoint

   Generated videos would be saved in path/to/checkpoint/animation

5. License

MIT