Deep self-attention network for facial emotion recognition

Human emotion detection is one of the major problems in computer vision. Human emotions consist of several sub-emotions which are difficult to classify into a specific class. In this proposed work, we have tried to classify the emotions into 6 basic categories (happy, sad, disgust, fear, surprise, anger) and neutral human emotions. We have proposed deep-learning framework which consist of CNN, ResNet and attention block which gives visual perceptibility to the network. The proposed model has greater applicability in real life facial emotion detection. The proposed model has achieved satisfactory result and has shown effective results on FER dataset.