To train the model from scratch -
- Download the Cohn-Kanade dataset.
- Create a folder named "data" in the project directory.
- And paste the dataset with the name "Dataset_images" and emotions with the name "Emotion" in the above created folder.
- Finally, open terminal in the project directory and then type in-
python train.py
After training or if you directly want to try the trained model -
python main.py
Dependencies -
- Numpy
- Pandas
- scikit-learn
- opencv
The emotions being classified are Neutral, Anger, Contempt, Disgust, Fear, Happy, Sad, Surprise
Currently the accuracy on the test data is 53.42% and live feed is not performing great.
The size of the dataset is not large enough to classify 8 different emotions, so probably train on less number of classes?
Currently under development