A Deep Learning Framework for Robust Facial Expression Recognition
This is the implementation of the proposed framework.
To run this you need to place the raw data, which is the SFEW dataset v2 into the direction data/dataset/raw_data
Then use the provided scripts to crop faces, split train and test set, and do data augmentation. Please specify correct target and source path.
The folder models are our networks. To run them, you need the below requirements:
You can also download the pre-processed dataset from here.
https://drive.google.com/file/d/1E5NITEV_i0DHsz8KLvPr66n4gV6JRdTY/view?usp=sharing
The voting mechanism is not integrated in the network. We manually predict the classes and process the output by jupyter notebook. This is shown in the folder notebooks
tensorflow-gpu
cv2
pytorch
scikit-learn