ABSTRACT

According to the United States Department of Education, Facial recognition technology has become increasingly popular for attendance tracking in educational institutions, offering a contactless and efficient alternative to manual methods. However, traditional systems face challenges such as the need for large datasets, retraining, and declining accuracy with more extensive databases. Advances in deep learning, like InsightFace, have helped overcome many of these challenges.

This study addresses a research gap by developing an automated attendance system that integrates real-time accurate identification with user-friendly interfaces. This system addresses privacy concerns and technical challenges in educational settings. The system uses InsightFace for feature extraction, Redis for real-time database management, and Streamlit for the user interface, providing a scalable and robust solution.

The system's empirical evaluation showed an accuracy and precision of 90% and a recall rate of 100%. While slightly below the benchmark set by the LFW dataset, the system's F1-Score of 94.74% indicates high reliability and validity.

These results represent a significant advancement in applying facial recognition systems in academic environments. The perfect recall rate, indicating the system's proficiency in identifying all present individuals, is a reassuring sign of its reliability for attendance systems. The F1-Score suggests a balanced precision-recall trade-off. While there is room for improvement in accuracy and precision, the developed system provides a strong foundation for deploying secure, efficient, and automated attendance systems in educational institutions.