Object Detection for Real-Time Malpractice Detection in Classrooms Using Computer Vision

Abstract

Ensuring academic integrity during examinations requires timely and effective detection of unauthorized items and suspicious behaviors. This project focuses on the object detection module within a larger system designed to identify students, invigilators, and prohibited items, such as mobile phones and notes, in classroom environments. A comparative analysis of detection models, including YOLOv8, YOLOv11, Fast R-CNN, and SSD, was conducted to evaluate key metrics. The results of different experiments revealed that YOLOv8 was the most successful model because of its high accuracy, occlusion-handling ability, and real-time processing capabilities. YOLOv8 performed better in real-world classroom scenarios by striking the ideal mix between speed and accuracy, making it a dependable option for object detection in dynamic environments. These results use YOLOv8's potential as the core component of a comprehensive system for real-time malpractice detection, ushering in integrity and fairness in exams. By having such advanced object identification methods embedded within monitoring systems, this approach ensures morality in AI-based solutions that emphasize equity and accountability in education.