Teacher Behavior Intelligent Recognition based on YOLO Framework and SPPF-Attention

Abstract

The behavior of teachers in the classroom has a crucial impact on students. In the past, experts mainly evaluated teachers’ performance in the classroom by observing their behaviors and providing ratings. This method is usually inefficient, time-consuming, and subjective. A teacher behavior intelligent recognition method was proposed in this paper. Based on the framework of YOLO V11, Attention mechanism and SPPF module was combined to form a new module SPPF-Attention which used to explore key features from feature layers of different sizes to recognize teacher behaviors in the classroom. Experimental results on SBC5_Teacher_Behavior dataset showed the proposed method obtained mAP value of 0.940 and had better recognition performance compared with other classic methods such as YOLO V8, YOLO V11, YOLO V12 and YOLO V13.