Improved YOLOv8 algorithm for student behavior detection in a smart classroom

In light of the advancements in educational technology, the smart classroom has emerged as a pivotal element in the contemporary educational landscape. Among these, student behavior detection technology is of particular importance for enhancing the quality of teaching and facilitating students' personal growth. In order to address the shortcomings of limited public datasets and low detection precision, we have created the SBC Dataset and proposed a student behavior detection algorithm, CCB-YOLOv8, which builds upon the YOLOv8 framework. Firstly, Coordinate Attention mechanism is introduced in the YOLOv8 backbone network to enhance the model's spatial perception. Secondly, the original UpSample operator is replaced by the CARAFE operator to augment the model's ability to capture key features. Finally, we introduce BiFPN in the Concat module to improve the precision and generalisation ability of the system. In comparison to YOLOv8, the precision of CCB-YOLOv8 has been enhanced by 4.3%, the recall by 5.6%, and the [email protected] and [email protected]:0.95 by 4.4% and 5.5%, respectively. These results demonstrate that the improved algorithm exhibits heightened accuracy and robustness.