International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 11 | Nov 2025
p-ISSN: 2395-0072
www.irjet.net
SMART PROCTOR – A COMPUTER VISION APPROACH TO ONLINE EXAM MONITORING Chaitra K J1 ,Abhishek G V2, Adarsh R Bongale3, Channabasappa Gowda B S4 ,Harshith K S5 1 Assistant Professor, Information Science and Engineering, Bapuji Institute of Engineering and technology,
Karnataka, India
2 Bachelor of Engineering, Information Science and Engineering, Bapuji Institute of Engineering and technology,
Karnataka, India
3 Bachelor of Engineering, Information Science and Engineering, Bapuji Institute of Engineering and technology,
Karnataka, India
4 Bachelor of Engineering, Information Science and Engineering, Bapuji Institute of Engineering and technology,
Karnataka, India
5 Bachelor of Engineering, Information Science and Engineering, Bapuji Institute of Engineering and technology,
Karnataka, India ---------------------------------------------------------------------***--------------------------------------------------------------------these technologies, this project introduces Smart Proctor, Abstract - Smart Proctor is an AI-powered online
an AI-driven remote exam monitoring solution that ensures secure, transparent, and real-time supervision. The system utilizes YOLOv8 for multi-person detection and object recognition, MediaPipe for accurate facial landmark tracking and gaze estimation, and facerecognition models for validating student identity throughout the exam. Together, these modules enable the detection of impersonation, additional persons in the frame, frequent head movements, screen diversion, use of prohibited devices, and suspicious background activity.
examination monitoring system developed to ensure academic integrity during remote assessments. The system uses a combination of deep learning and computer vision technologies such as YOLOv8, MediaPipe, OpenCV, and Face Recognition models to detect impersonation, additional persons, gaze deviation, and suspicious audio activity. A Flask-based backend processes webcam and audio streams in real time, while a MySQL database securely stores violation logs and session histories. The system achieved more than 97% face-matching accuracy and maintained an average processing latency of 1.3 seconds during testing. Smart Proctor provides a reliable, scalable, and automated proctoring solution suitable for academic institutions.
To support continuous monitoring, Smart Proctor integrates a Flask-based backend that processes live webcam feeds and audio in real time. Each detected event is logged into a MySQL database with timestamps, confidence levels, and violation categories, creating a complete behavioral record for exam administrators. In addition, background voice activity is monitored to identify instances of verbal prompting or external assistance. Unlike traditional systems that rely solely on manual supervision, Smart Proctor offers automated decision-making capabilities that reduce human involvement while improving detection accuracy.
Key Words: Online Exam Monitoring, Smart Proctoring, YOLOv8, MediaPipe, Deep Learning, Computer Vision, Face Recognition, Flask
1.INTRODUCTION With the increasing shift toward online learning and remote examinations, maintaining academic integrity has become a significant concern for educational institutions. Traditional in-person invigilation is no longer feasible in virtual exam environments, resulting in a rise in impersonation, unauthorized assistance, and other forms of malpractice. Although many institutions initially relied on manual video monitoring or basic webcam-based supervision, these methods are highly inefficient, prone to human error, and incapable of detecting subtle or realtime violations. The absence of standardized monitoring mechanisms often leads to unresolved malpractice, affecting grading fairness and overall academic standards.
By combining advanced AI models with a scalable webbased architecture, the proposed system provides a reliable, privacy-aware, and efficient solution for conducting secure remote examinations. Its modular design allows institutions to customize sensitivity levels, extend features, and deploy the system across various online learning platforms. Smart Proctor represents a modern approach to protecting academic integrity and ensuring fair evaluation in the era of digital education.
2. PROPOSED SYSTEM
Recent advancements in artificial intelligence, deep learning, and computer vision have enabled the development of automated proctoring systems that can mimic and even surpass human invigilators. Leveraging
© 2025, IRJET
|
Impact Factor value: 8.315
The integrated system combines AI-powered exam monitoring with automated violation logging, creating a complete end-to-end online invigilation pipeline.
|
ISO 9001:2008 Certified Journal
|
Page 613