<![CDATA[This study aims to develop a student learning activity monitoring system based on deep learning and computer vision using YOLOv12n, DeepSORT, MediaPipe FaceMesh, and the Fuzzy Mamdani method. The study was conducted on eighth-grade students of SMP Swasta Deli Murni Suka Maju using classroom learning activity videos that had undergone data selection and preprocessing. The proposed system was designed to detect students, track their identities, and analyze their attention levels automatically and in real-time based on eye conditions and head position. YOLOv12n was employed for student detection, DeepSORT for identity tracking, MediaPipe FaceMesh for extracting facial features, including the Eye Aspect Ratio (EAR) and head position, while the Fuzzy Mamdani method was utilized to classify students' attention levels. System evaluation was performed by comparing the prediction results with manually annotated ground truth and was assessed using the Confusion Matrix, accuracy, precision, recall, and F1-score. The experimental results demonstrate that the proposed system is capable of performing multi-student detection, identity tracking, and automatic attention level analysis. The evaluation achieved an accuracy of 90.20%, indicating that the integration of YOLOv12n, DeepSORT, MediaPipe FaceMesh, and the Fuzzy Mamdani method provides reliable performance for classifying students' attention levels in a smart classroom environment.]]>
Copyrights © 2026
