Emotion Recognition from Static Human facial Images

This was our project for CS771A - Machine Learning Techniques,Fall 2016. We implemented various supervied classifiers such as naive bayes, Kernelized SVMs etc. for emotion classification on human facial Images from the Emotion Recognition in the Wild Challenge 2016 dataset.
For supervised classifiers, we engineered suitable feature vectors using Google Cloud Vision API and annotated the ground truth in a semi-supervised fashion using Cloud Visio API and human supervision. Futher we compared the performance of traditional classification approaches against results from pre-trained Convolutional Neural Networks in our project’s final report. Also, the slides for the presentation are available here.