Smart Attendance Face Recognition System (SAFRS): A Deep Learning Approach to Automated Biometric Attendance Management

This paper presents the design, implementation, and evaluation of the Smart Attendance Face
Recognition System (SAFRS), an enterprise-grade AI-powered platform for automated attendance
management in educational institutions and corporate environments. SAFRS integrates Multi-task
Cascaded Convolutional Networks (MTCNN) for real-time face detection with FaceNet embeddings and
Support Vector Machine (SVM) classification to achieve a top-1 recognition accuracy of 99.8%, a False
Acceptance Rate (FAR) of 0.02%, and sub-second end-to-end detection latency of 221 ms on standard
hardware. The system eliminates proxy attendance through a parallel liveness detection module
requiring a minimum liveness score of 0.85, and supports simultaneous recognition of 50 or more
individuals in a single video frame. The architecture follows a modular four-tier design encompassing a
hardware capture layer, an AI processing pipeline, a RESTful backend with WebSocket support, and a
React.js-based administrative dashboard. Security design complies with GDPR Article 9, FERPA, and
ISO/IEC 27001.