Classroom Monitoring System For Exam and Behavior Using Deep Learning

Abstract

The increasing adoption of digital examination systems and smart classroom environments has created a demand for reliable automated monitoring solutions that ensure academic integrity. This work presents a web-based classroom monitoring and online proctoring system that integrates deep learning and computer vision techniques for real-time supervision. The system is developed using Flask for backend operations, OpenCV for video processing, YOLOv8 for object detection, MediaPipe for facial analysis, and PostgreSQL for structured data storage.The proposed solution continuously analyzes live video streams to identify potential malpractice indicators, including the presence of mobile phones, books, multiple individuals, absence of a student, abnormal head movements, poor lighting conditions, and suspicious motion patterns. Upon detecting violations that exceed predefined thresholds, the system captures evidence images, logs events with timestamps, and triggers email alerts to administrators.A role-based interface enables administrators to manage students, subjects, examinations, and malpractice records, while students can securely access and participate in scheduled assessments. Additionally, the system supports offline classroom monitoring, extending its usability beyond online examinations. By combining real-time analytics with centralized data management, the system enhances monitoring accuracy, reduces reliance on manual invigilation, and ensures traceable and consistent supervision.

DOI: https://doi-ds.org/doilink/04.2026-67656811