Utilising the YOLOv3 Algorithm for the Student Posture Recognition System in Classroom Settings

Effective student-teacher interaction helps transfer knowledge, clarify concepts, and create a conductive learning environment. The effectiveness of the interaction can be seen through students’ behaviour and various factors, such as pos- ture and gestures in the classroom. However, educators face significant difficulties in tracking each student’s performance and behaviour during class. Therefore, this study focuses on student posture recognition in classroom settings, which is essential for monitoring student behaviour and engagement during lectures. The proposed system utilises the YOLOv3 machine learning model for real-time detection. A dataset of student postures was collected from Google Images, and the data was used to train a deep neural network. The model was then tested on classroom images and compared to manual annotations. The results showed that the model can accurately recognise student postures with high precision, recall, F1-score, and mean average precision (mAP), achieving an average precision of 88%, recall of 89%, F1-score of 88%, and mAP of 95.20%. The real-time processing capability of YOLOv3 allows for immediate posture detection during lectures in a classroom; this may help educators monitor student behaviour and engagement.