Deep ConvNet for Facial Emotion Recognition

Abstract

Facial expression recognition has made great progress over the last two decades. The growth in the use of deep learning has contributed significantly to this advance. In this work, we proposed a Deep ConvNet architecture for facial expression recognition based on the RaFD dataset. Results show that the best setup is the combination with different parameters like the convolution layer number, RELU function activation in hidden layers, Adam optimizer, 2 fully connected layers…etc. The best architecture gives over 97% accuracy which is promising compared to the state-of-the-art results and confirms the effectiveness and robustness of DCNN with batch-normalization.