An Novel End-to-end Network for Automatic Student Engagement Recognition

Abstract

Automatically recognizing and improving students engagement is conducive to improving students academic performance and less dropout rate. After analyzing the advantages and disadvantages of some methods in the field of automatic recognition of students engagement, in this paper, a novel end-to-end student engagement recognition network is proposed. We innovatively improve and optimize the excellent model Inflated 3D Convolutional Network (I3D) used in the field of action classification and apply it to the field of automatic students engagement recognition. In the end, we have achieved ideal results in the student engagement recognition dataset DAiSEE, with an accuracy of 98.82 % in binary engagement classification, 4.42% higher than the benchmark.