Student Behavior Detection in the Classroom Based on YOLOv10 and Edge Computing

Detecting students' classroom behaviors from instructional videos is important for instructional assessment, analyzing students' learning status, and improving teaching quality. However, most contemporary methods for recognizing student behavior in classrooms rely on GPU servers, which impose significant network load, compromise security, and struggle to achieve real-time detection. To address these challenges, this study proposes a novel method for student classroom behavior detection based on YOLOv10 and edge computing. Firstly, the Multi-Scale Attention Model (MSAM) is incorporated into the YOLOv10 network architecture to enhance the model's detection accuracy, leading to the formulation of the YOLOv10-MSAM network structure. Secondly, the YOLOv10n-MSAM model is optimized using TensorRT and deployed on the Jetson Orin NX edge device. During deployment, CUDA acceleration is applied to both the preprocessing and postprocessing stages. In the experiments, the YOLOv10-based model achieved a mean Average Precision (mAP) of 83.7% on the given dataset with an Intersection over Union (IoU) threshold of 0.5. Upon integrating the Multi-Scale Attention Model (MSAM) into the YOLOv10 architecture, the proposed method demonstrated a significant enhancement in performance, achieving mAP0.5 and mAP0.5:0.95 scores of 87.0% and 67.0%, respectively. Additionally, the inference speed of the TensorRT-optimized model is nearly 40 times faster. Finally, after a series of deployment optimizations, the YOLOv10n-MSAM model maintains an average processing frame rate of over 50 frames per second on the Jetson Orin NX edge device, meeting the requirements for real-time classroom behavior detection.