Improved YOLOv11 for Classroom Behavior Detection and Student Counting

Currently, the application of Artificial Intelligence for classroom student behavior detection and attendance has become an effective way to improve the efficiency of manual supervision. However, existing research has not yet effectively resolved the challenges of multi-scale variations, small target detection, and occlusion in classroom scenarios. Additionally, models trained on public datasets often perform poorly when applied to real-world scenarios. To address the aforementioned challenges, this paper proposes an improved YOLOv11 model by introducing the Enhanced Visual Attention Block (EVCBlock) and Multiscale Attention Aggregation (MSAA) modules. These improvements are designed to improve the model's detection accuracy for small and occluded targets, while simultaneously optimizing the efficiency of multi-scale feature fusion. The experimental results indicate that the [email protected] of the improved model in classroom behavior detection reaches 91.51%, marking a 1.09% improvement over the YOLOv11 model. Similarly, in student counting, the precision of the improved model reaches 91.96%, reflecting a 0.47% enhancement compared to the YOLOv11 model. Through testing on real classroom environment datasets, it is verified that the proposed method can accurately detect student behavior with varying levels of occlusion in complex classroom environments, effectively reduce misdetections and omissions, and demonstrate strong robustness and potential for practical applications.