Attendance System Based on Face Recognition

Facial recognition has established itself as a key technology with multiple applications in security, authentication, and access control. This paper presents the development of a facial recognition-based attendance registration system. It uses the InsightFace library (buffalo_sc model) to extract 512-dimensional vectors and cosine similarity for real-time identification. The process includes employee registration, image capture, embedding comparison against a Redis database, and a front-end using Python Streamlit. During the initial evaluation phase with 30 participants, the system achieved a recognition success rate of 96.7% and an average recognition time of 3.0 s, with a maximum camera initialization time of 10.0 s. These results demonstrate high accuracy and an efficient system for automating attendance registration processes.