AI-DRIVEN EXAMINATION HALL MONITORING SYSTEM USING DEEP LEARNING FOR REAL-TIME MALPRACTICE DETECTION
Santhosh C, Adhithyaa N, Lokesh S
Keywords: YOLOv8, Computer Vision, Deep Learning, Object Detection, Automated Proctoring, Real-Time Detection, Academic Integrity, Examination Supervision.
Abstract
Ensuring the integrity of examinations remains a salient concern in modern educational contexts despite the advancement of technology in most other areas. Traditional human proctoring has drawbacks such as time lapses that promote fatigue, inconsistency, and the inability to monitor persistently. This is an innovative solution: an AI-powered examination hall monitoring system utilizing the YOLOv8 deep learning algorithm for real-time malpractice detection. The system is designed to detect the variability of examination malpractices such as mobile usage, looking at other people's paper, suspicious hand movements, and utilization of unauthorized materials during an examination. Our implementation uses live video feed processing by techniques like computer vision and employs a custom-trained YOLOv8 model to identify suspicious behaviors. The system incorporates an automated alert mechanism that instantly notifies examination authorities via email when potential malpractices are detected. Through rigorous testing with a diverse dataset of examination scenarios, our model achieved significant accuracy in detecting multiple categories of malpractice behaviors. Our results show the ability of the system to discern normal examination behavior from suspicious activities with real-time processing capability and minimal false positives. This study contributes to the advancement of automated examination supervision technologies as well as presents a scalable solution to support academic integrity in modern educational settings.