Accurate facial recognition is essential in modern classroom environments, enabling automated attendance tracking and real-time monitoring of student participation. However, classroom settings present unique challenges, including low-resolution images caused by distance, varied lighting conditions, and occlusions, which significantly reduce identification accuracy. While previous approaches often employed super-resolution methods to address these issues, they required high computational resources and offered suboptimal accuracy. This study proposes using YOLOv8 to enhance face detection and recognition specifically tailored for classroom conditions. Experiments were conducted with four YOLOv8 variants—YOLOv8-S, YOLOv8-M, YOLOv8-L, and YOLOv8-X—in real classroom settings involving 40 students within a 6 m x 5 m space. The results demonstrate that YOLOv8-X delivered the best performance, achieving 92% precision, 88% recall, and an mAP50 of 95%, proving highly effective for detecting students in challenging classroom scenarios. YOLOv8-L closely followed with 94% precision and 84% recall. In contrast, YOLOv8-M and YOLOv8-S showed limited effectiveness, with YOLOv8-S achieving only 82% precision and 70% recall. These findings highlight the suitability of YOLOv8-L and YOLOv8-X for addressing the complex challenges of classroom environments, providing robust solutions for improving facial recognition accuracy and efficiently automating classroom management systems.
Ananda et al. (Fri,) studied this question.