Student Facial Expression Recognition Based On Multi-scale Residual Network

The monitoring of student learning status is a hot topic in the field of visual analysis such as intelligent classroom and digital teaching. Its essence is the students’ facial expression recognition. In this paper, we proposed an expression recognition method based on multi-scale residual network to solve the problems of different degrees of occlusion in the classroom. The model realizes the fusion of multi-scale features for expression recognition by integrating different scale features, and effectively inhibits the influence of partial occlusion in the classroom scene. To verify the effectiveness of the model, the data-set of students’ facial expressions was established based on the actual classroom scene. The average accuracy of the test reached 93.6%. The results of the experiments show that the propose model could accurately identify students’ facial expressions in class.