Expression Recognition using YOLO and Shallow CNN Model

In this digital era, identifying human facial expressions and responding accordingly is an emerging need. On the other hand, similar data can be used for surveillance activities and relative crime detection. Human Facial Expression Recognition (HFER) based on human face expression variations in real-time is proposed in this paper. Here two CNN- models are cascaded to produce the facial expression recognition output. YOLO V5 is used for people detection and custom trained CNN model is used for expression recognition. The suggested model provides better accuracy of 95.57% for seven different facial expressions than the existing models.