Real-time Detection of Students’ Cellphone Usage During Lectures and Behavioral Analysis

This research paper explores the application of Computer Vision in the realm of classroom management by addressing the issue of cellphone usage during lectures among students. We present a method for cellphone detection in classrooms during lectures, which can be used to reflect on students’ distraction levels. Using CCTV camera recordings from college classrooms, we created two datasets: one comprising standard lecture recordings and the other capturing an empty classroom environment with only tables visible. Our system utilizes two fine-tuned YOLOv8 models, trained on these datasets, to detect and track students’ cellphone usage patterns. The table detection model maps out all the tables present in the classroom and the cellphone detection model detects cellphones and their users. The cellphone detection model was trained using three variations of YOLOv8: small (s), medium (m), and large (l). The best performing model was YOLOv8m, trained for 55 epochs, achieving an mAP50 of 89.7%. The integration of computer vision technology into classroom management not only supports educators in enforcing cellphone usage policies but also provides valuable insights for optimizing teaching strategies. Our system facilitates behavior analysis by enabling continuous monitoring of students’ cellphone usage over time. Preliminary evaluation of the models shows promising results in accurately detecting and tracking cellphone usage during lectures.