An Integrated Model for On-Site Teaching Quality Evaluation Based on Deep Learning

Abstract

During on-site teaching for university students, the level of concentration of every student is an important indicator for the evaluation of teaching quality. Traditionally, teachers rely on subjective methods for observing students’ learning status. Due to the volume of on-site crowds, teachers are unable to stay on top of the learning status of each student. Meanwhile, because of the subjective evaluation, the results would not be precise. With the fast development of artificial intelligence and machine learning, it is possible to adopt deep learning technology to achieve scientific evaluation of the classroom teaching quality. This paper proposes an integrated evaluation model based on deep learning technology, incorporating YOLOX model, Retinaface model, and SCN model. Among which, YOLOX model is used to detect the area of the students’ upper body, Retinaface model is adopted to assess the head-up rate, and SCN model is used to recognize the facial expression. The experimental results have shown that our model can achieve 93.1% object detection accuracy, more than 85% face recognition accuracy, and 87.39% expression recognition accuracy. We further develop a model to use the combination of head-up rate and facial expression scores to jointly evaluate classroom teaching quality. Five teaching professors’ evaluations of our classroom video images confirmed that our proposed model is effective in objectively evaluating the on-site teaching quality.