Volume 1 - Issue 2, May - June 2026

Maintaining discipline during examinations becomes difficult when a limited number of invigilators are responsible for monitoring large classrooms continuously for long durations. In many situations, suspicious activities such as looking toward neighboring students, passing answer sheets, checking mobile phones, or communicating through gestures may remain unnoticed during manual supervision. This paper presents an AI-based smart examination invigilation system developed using deep learning and computer vision techniques for automated classroom monitoring. The proposed framework uses the YOLOv8n object detection model together with live webcam or CCTV video streams to identify suspicious student activities during examinations in real time. One of the important parts of this work was the preparation of a custom classroom dataset using recordings collected under realistic examination conditions. Twenty-seven classroom videos were captured under different seating arrangements and lighting conditions. The recorded videos were processed using Python and OpenCV libraries to extract image frames for training. During preprocessing, several extracted frames were found to be repetitive or blurred because consecutive frames contained very little variation. To improve dataset quality, additional preprocessing operations such as frame skipping, duplicate frame removal, blur filtering, and annotation correction were performed manually. The final dataset contained 870 annotated images prepared using Roboflow for training and evaluation. Experimental observations showed that careful preprocessing and annotation refinement improved overall detection consistency during classroom monitoring. The proposed framework demonstrates how AI-assisted surveillance systems can support educational institutions in improving examination transparency and automated monitoring.

Abhishek M, Gagan M, G Surya, Gladson K, "AI-Based Smart Examination Invigilation System Using Deep Learning" International Journal of Computer Science and Engineering Development, V1(2): Page(19-22) May-June 2026. ISSN: 3139-0862. www.ijcsed.com. Published by Scientific and Academic Research Publishing.