Automated Student Activity Recognition Using YOLOv11n

The proposed work aims to develop an automated approach for recognizing student activities in classrooms using a YOLOv11n-based neural network model. Traditional manual observation methods are often time-consuming and limited in scope, prompting the need for more efficient and accurate alternatives. The growing demand for intelligent classroom monitoring highlights the importance of such approaches for improving learning analytics and classroom management. To address the issue, the model was trained on a diverse dataset that includes a wide range of student postures and interactions within the classroom environment. The dataset was enriched through data augmentation, expanding from 2262 to 5428 images, which contributed to greater diversity and robustness. The YOLOv11n algorithm was employed for model training, yielding outstanding results. With a mean average precision ($\mathbf{m A P}$) of $\mathbf{9 9. 5 \%}$, the model outperformed alternative approaches, including YOLOv7 and Faster R-CNN, in both precision (99.6%) and recall (99.8%). The results validate the capability of YOLOv11n for efficient realtime monitoring of student activities in classrooms, providing a robust solution for improving educational analysis and tracking student engagement.