• Abhishek Narayana Sharma
• Ananya
• Anusha
• Deeksha
• Prof. Aravind Naik

Automated exam surveillance using computer vision can greatly enhance the integrity of examinations by detecting suspicious student gestures in real time. This paper presents an intelligent architecture that uses multi-camera CCTV feeds, deep learning-based gesture recognition (e.g. Open Pose, YOLO-Pose), and machine learning classifiers to flag cheating behaviors. The system integrates pose estimation, object detection, and temporal analysis to recognize acts such as passing notes or looking at nearby papers. In simulated experiments, such methods have achieved high accuracy (over 90%) in distinguishing cheating from normal behavior. The proposed approach is benchmarked on public exam behavior datasets (e.g. the CUI-EXAM dataset with ~4000 labeled exam images) and uses metrics like accuracy, precision, and recall for evaluation. We discuss challenges in deploying this system (computational load, camera coverage) and ethical considerations (privacy, bias). The architecture is illustrated in Fig.1. Keywords: exam integrity, real-time invigilation, gesture recognition, computer vision, smart proctoring.

Real-time invigilation, Cheating detection, Gesture recognition, Computer vision, Deep learning

"Intelligent Exam Surveillance: Gesture Based Real - Time Cheating Detection", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.12, Issue 7, page no.467-471, July-2025, Available :http://www.jetir.org/papers/JETIRGX06088.pdf

"Intelligent Exam Surveillance: Gesture Based Real - Time Cheating Detection", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.12, Issue 7, page no. pp467-471, July-2025, Available at : http://www.jetir.org/papers/JETIRGX06088.pdf