An AI-Aided Student Classroom Behavior Detection Algorithm Based on Classroom Images

In focus on the problems of low detection accuracy, high density of detection targets, and serious occlusion in the current student classroom behavior detection, this paper proposes an improved student classroom behavior detection algorithm RB-YOLO11 based on YOLO11. Firstly, the RFAConv module is introduced to overcome the limitations of the convolutional kernel caused by the sharing of image parameters in different regions of the processing, so as to improve the model's ability to recognize and learn from complex scenes and patterns; and then the BiFormer attention mechanism is added to significantly improve the model's accuracy. The testing experiments are conducted with the students' classroom behavior dataset. The experimental results show that compared with YOLO11, the mAP of RB-YOLO11 is improved by 2.4%, Precision by 3.3% and Recall by 1.2%, which indicates that the algorithm has higher detection precision and stronger generalization ability, and is capable of performing the task of detecting students' classroom behavior.