Student Engagement Detection Using YOLOv8m and MediaPipe

This technical report presents a novel method for detecting student engagement in classroom settings by utilizing a combination of You Only Look Once Version 8 Mini (YOLOv8m) and MediaPipe. The study addresses the limitations of previous methods that relied on outdated object detection techniques and cumbersome manual dataset creation. The proposed method demonstrates significant improvements in accuracy and speed, achieving a cross-entropy loss of 0.40 compared to the baseline of 0.60, and operates approximately 16 times faster in pose detection data collection. The report outlines the methodology employed, including the use of YOLOv8m for object detection and MediaPipe for keypoint detection, and compares the results against a baseline method using YOLOv4 and OpenPose. Statistical significance of the results is assessed through a paired t-test. The findings indicate that the proposed method not only enhances detection accuracy but also streamlines the data collection process, making it a promising approach for monitoring student engagement in educational environments.