A Two-Stage, Object-Centric Deep Learning Framework for Robust Exam Cheating Detection

2026-04-17 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Computer Vision & Pattern Recognition · Depth: Advanced, quick

Summary

A new two-stage deep learning framework has been developed to enhance exam cheating detection, addressing the inefficiencies and costs of human invigilation. This system integrates object detection with behavioral analysis, utilizing a YOLOv8n model to localize students in exam-room images. Subsequently, cropped student regions are classified by a fine-tuned RexNet-150 model to identify normal or cheating behaviors. The framework was trained on a dataset of 273,897 samples from 10 sources, achieving 0.95 accuracy, 0.94 recall, 0.96 precision, and a 0.95 F1-score. This represents a 13% improvement over a baseline accuracy of 0.82 in video-based cheating detection. With an average inference time of 13.9 ms per sample, the system is designed for robustness and scalability in large-scale deployments, while also incorporating ethical considerations by delivering private results to students.

Key takeaway

For institutions seeking to implement or upgrade AI-powered exam proctoring, this framework offers a robust, scalable, and ethically designed solution. You should consider adopting a two-stage object-centric approach, leveraging models like YOLOv8n and RexNet-150, to achieve high accuracy and efficient inference. Ensure your system includes private result delivery to students to address ethical concerns and promote reflection.

Key insights

A two-stage deep learning framework improves exam cheating detection with high accuracy and ethical private feedback.

Principles

• Combine object detection with behavioral classification.
• Prioritize ethical delivery of monitoring outcomes.

Method

The method uses YOLOv8n for student localization, followed by a fine-tuned RexNet-150 to classify cropped regions as normal or cheating behavior, trained on a large multi-source dataset.

In practice

• Deploy YOLOv8n for initial object localization.
• Fine-tune RexNet-150 for specific behavior classification.
• Implement private result delivery to students.

Topics

• Exam Cheating Detection
• Deep Learning Frameworks
• YOLOv8n
• RexNet-150
• Computer Vision

Best for: Computer Vision Engineer, Research Scientist, AI Scientist, Machine Learning Engineer, AI Engineer

Related on AIssential

• A Two-Stage, Object-Centric Deep Learning Framework for Robust Exam Cheating Detection — A two-stage, object-centric deep learning framework significantly improves exam cheating detection accuracy and scalability.
• YOTOnet: Zero-Shot Cross-Domain Fault Diagnosis via Domain-Conditioned Mixture of Experts — YOTOnet enables zero-shot cross-domain fault diagnosis in mechanical equipment using a domain-conditioned mixture of experts.
• Teacher-Student Structure for Domain Adaptation in Ensemble Audio-Visual Video Deepfake Detection — A teacher-student framework enhances ensemble audio-visual deepfake detection models for robust generalization across diverse, unseen domains.
• On the Study of Biometric Spoofing Detection using Deep Learning — Biometric spoofing detection benefits from efficient deep learning models like MobileNetV2, but generalization remains a challenge.
• [P] Open-source ML homeworks with auto-tests - fundamental algorithms from first principles — Building fundamental ML algorithms from scratch with automated testing enhances deep understanding and provides immediate feedback.
• Deep-testing: the case of dependence detection — Deep-testing uses neural networks for hypothesis testing, outperforming traditional methods in dependence detection.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.