This is real time facial expression detector where the ML model is trained using CNN via TensorFlow. The data set used was from the FER-2013 dataset from Kaggle (https://www.kaggle.com/msambare/fer2013). The data set maybe biased towards certain emotions and thus to have a equal number of training images data augumentation was used. You can find my data augumentation script for images in my other repo.
Here are some results ,
The accuracy for validation is around 0.65 as shown below
