Classroom Engagement through Real-time Student Attention Analysis using YOLO V8

Student engagement is crucial for the realization of learning outcomes and improvement in academic performance. It is not easy to monitor students to ascertain their attention, especially for a large class, in a traditional classroom setting. This work leverages modern machine learning and computer vision by using the state-of-the-art object detection model, YOLO v8, to monitor student engagement in real-time. The system processes posture and facial expressions into key engagement indicators that provide actionable insights for teachers to further refine instructional methods to improve interaction in the classroom. The model was thus trained on the engagement levels captured by the frames of the annotation video, which includes "Focused," "Distracted," "Leaning Forward," and "Looking Away." YOLO v8 model performance evaluation showed promising results, where precision, recall, and mean average precision proved its accuracy in classifying the engagement indicators. Real-time testing also confirmed that the model could meet the frame rate and latency requirements for a classroom environment. Error analyses showed a deficiency in identifying subtle facial expressions and gaze directions in cases of poor light conditions- outlining topics for further refinement. Overall, this work has been able to present the possibility of embedding systems with YOLO v8 into the classroom context to help enhance teacher capacity and student engagement in learning effectively. Further research is recommended to apply the model in a wide environment of learning with the hope of wider educational impacts.