Lightweight Face Recognition System for Automated Attendance Tracking Under Data-Constrained Environments

Traditional attendance tracking in educational institutions relies on labor-intensive manual methods, including roll calls and paper-based registration, which are inherently inefficient and susceptible to fraudulent practices. This study presents an automated, contactless attendance system leveraging advanced computer vision techniques optimized for resource-constrained deployment scenarios. The proposed framework integrates YOLOv8n-face for real-time face detection with a fine-tuned ArcFace-R100 model for robust identity verification, specifically designed for edge device deployment. To address the fundamental challenge of limited training data prevalent in educational settings, we implement a comprehensive data augmentation strategy that enhances model robustness against variations in illumination, facial pose, and partial occlusions. Experimental validation on the Labeled Faces in the Wild (LFW) benchmark demonstrates a recognition accuracy of 96.7% with real-time inference at 11.74 FPS. Real-world deployment in classroom environments confirms the system’s practical effectiveness, scalability, and security, establishing it as a viable modern alternative to conventional attendance management approaches.