Recognition Model Construction and Analysis of College Students' Classroom Behavior

In college classroom teaching, students' classroom behaviors serve as critical evidence for teachers to evaluate learning outcomes and provide personalized instruction. Traditional classroom behavior data suffer from limited dimensions and lack coverage of low-resolution classroom scenarios. This paper constructs a College Student Classroom Behavior dataset containing 8,845 classroomimages with 36, 607 annotated labels across four student behaviors and achieves the college student behavior recognition task based on YOLOv8. Experimental results demonstrate that the trained model achieves high recognition accuracy. Descriptive statistical analysis of collected classroom behavior data reveals regular behavioral changes among students at different grades, offering great significance for teaching activities.