Apache MXNet GLUON implementation of state of the art FER+ paper by Barsoum et. al. for facial emotion recognition. You can see the final demo at - http://bit.ly/mxnet-fer. In this implementation, we use majority voting (MV) technique illustrated in the paper and achieve 81.25% validation accuracy.(paper achieves 83.85%) We export the trained facial emotion recognition model and learn how to deploy at scale in production using MXNet Model Server.
By end of this session, you will learn about Apache MXNet and its powerful and easy to use GLUON interface in Python, basics of Convolutional Neural Network (CNN), Training a state of the art CNN model for Facial Emotion Recognition, Productionizing deep learning models at scale using MXNet Model Server.