Genetic Algorithm for feature subset selection.
Fitness for individuals (feature subset) defined using the mean accuracy for classifying emotions. Classification of emotions using various classifiers on the CK+ dataset and OpenFace.
Graphical User interface for entering csv path, loading the pickled trained model for instant classification on test data.
Logistic Regression - Mean accuracy 93.156 Neural Net with 2 hidden layers and one dense output layer - Mean Accuracy 91.331 Selects best feature subset of 434 features out of ~750 features generated from OpenFace.
Run the main.py file with all other files in the same directory