Real-Time Detection of Student Drowsiness in Online Learning Environments Using YOLO11

As online learning continues to grow in popularity, issues such as decreased student engagement and increased drowsiness have become common, often resulting in lower learning outcomes compared to offline learning. One critical challenge in this context is how to effectively monitor student participation in real-time situations. This study presents a real-time, image-based drowsiness detection system for online learning environments. It uses YOLO11 as it is currently one of the newest detection models and achieves high performance in object detection tasks. The system is trained and evaluated using the Kaggle Drowsiness Detection dataset, demonstrating consistent accuracy across various scenarios. Our model processes webcam video feeds and analyzes each frame to categorize the student’s condition as microsleep, yawning, or neutral. To improve accuracy, the system monitors behavioral patterns in consecutive frames and triggers alerts if thresholds are met. This approach ensures timely intervention to mitigate the effects of drowsiness. The YOLO11 model achieved an overall mean Average Precision (mAP@50) of 98%, demonstrating its effectiveness in identifying drowsiness-related behaviors. Additionally, the real-time implementation maintained low latency, with preprocessing, inference, and postprocessing times totaling less than 35 milliseconds per frame. This study underscores the potential of AI-driven systems to improve online learning by addressing challenges related to student engagement and drowsiness.