Recognition and Identification of College Students" Classroom Behaviors through Deep Learning

Abstract

Recognizing and managing college students" classroom behavior in a timely manner is of great help in improving teaching quality and strengthening classroom management. This paper builds a model based on the You Only Look Once Version 5 Small (YOLO v5s) algorithm using deep learning to detect and identify college students" classroom behaviors. The LabelImg annotation tool was used to process the dataset images, and the labeled dataset was the input for the object detection model to recognize college students" classroom behaviors. Although the precision, recall, mean average precision (mAP), and detection speed of the YOLO v5s model were slightly lower with large classroom densities, compared to medium classroom densities, the difference was negligible. At the same time, the mAP values of the proposed model under three different intersection-over-union thresholds were higher than the single shot multibox detector and regionbased convolutional neural network models, reaching 95.8, 94.3, and 92.9. This paper proves that YOLO v5s can effectively and accurately recognize classroom behavior in real time.