Dataset Used: FER2013 kaggle challenge dataset
The model has been trained on google colab and accuracy obtained on the testing dataset is 64.5%. Tuning the hyperparameters can give better accuracy. The model detects 6 emotions pretty accurately: Neutral, Angry, Fear, Surprise, Sad and Happy. It faces difficulty in detecting Disgust emotion. I am currently working on resolving this problem.
Training has been done on Colab GPU. Link to the colab notebook: https://colab.research.google.com/drive/1FujiF_-i7Jtaz8EWqhhWjxHXSnX5p4SR?usp=sharing
