pip install -r requirements.txt

data/
├─ RAF-DB/
│  ├─ basic/
│  │  ├─ EmoLabel/
│  │  │  ├─ images.txt
│  │  │  ├─ image_class_labels.txt
│  │  │  ├─ train_test_split.txt
│  │  ├─ Image/
│  │  │  ├─ aligned/
│  │  │  ├─ aligned_224/  # reagliend by MTCNN

data/
├─ AffectNet/
│  ├─ affectnet_info/
│  │  ├─ images.txt
│  │  ├─ image_class_labels.txt
│  │  ├─ train_test_split.txt
│  ├─ Manually_Annotated_Images/
│  │  ├─ 1/
│  │  │  ├─ images
│  │  │  ├─ ...
│  │  ├─ 2/
│  │  ├─ ./

% (1) images.txt:
idx | imagename
1 train_00001.jpg
2 train_00002.jpg
.
15339 test_3068.jpg

% (2) image_class_labels.txt:
idx | label
1 5
2 5
.
15339 7

% (3) train_test_split.txt:
idx | train(1) or test(0)
1 1
2 1
.
15339 0

python3 train_fer_first_stage.py \  
--dataset ${DATASET} \ 
--data-path ${DATAPATH}

python3 train_fer_second_stage.py \  
--dataset ${DATASET} \  
--data-path ${DATAPATH} \  
--ckpt-path ${CKPTPATH}

bash stage1.sh

bash stage2.sh

python3 train_fer_second_stage.py \ 
--eval \
--dataset ${DATASET} \       # dataset name
--data-path ${DATAPATH} \    # path to dataset
--ckpt-path ${CKPTPATH} \    # path to first stage ckpt
--eval-ckpt ${EVACKPTPATH}   # path to second stage ckpt

@ARTICLE{Zhou2024CEPrompt,
  author={Zhou, Haoliang and Huang, Shucheng and Zhang, Feifei and Xu, Changsheng},
  journal={IEEE Transactions on Circuits and Systems for Video Technology}, 
  title={CEPrompt: Cross-Modal Emotion-Aware Prompting for Facial Expression Recognition}, 
  year={2024},
  doi={10.1109/TCSVT.2024.3424777}
}