Pytorch-RAF-Resnet
A Pytorch Implementation of Facial Expression Recognition On RAF Dataset based on Resnet, I got 86.75% percision on 2 GPUs, you may get better results if you want to try more pretrained models.
Denpendency
pip install -r requirements.txt
Data
datasets/
└── raf-basic
├── EmoLabel
│ └── list_patition_label.txt
└── Image
└── aligned
list_patition_label.txt
data format is:
train_00001.jpg 5
train_00002.jpg 5
train_00003.jpg 4
train_00004.jpg 4
......
And the images are in aligned directory.
Training
python mainpro_RAF.py
Reference
https://github.com/WuJie1010/Facial-Expression-Recognition.Pytorch