Facial Recognition-Based Attendance Monitoring System for Non-Teaching Employees of St. Clare College of Caloocan

Abstract

The administrative operations of educational institutions require precise, reliable, and secure attendance tracking to maintain workforce accountability and ensure payroll integrity. Traditional manual logbook frameworks are vulnerable to transcription errors, timestamp manipulation, structural damage, and proxy attendance ("buddy punching"). This study presents the design and implementation of an offline-first, facial recognition-based attendance monitoring system optimized specifically for the 14 non-teaching employees of St. Clare College of Caloocan. The system incorporates the face-api.js library for client-side edge biometric computation, a secure PHP backend hosted via Apache, and a local MySQL infrastructure for secure record storage. Employing a mixed-methods developmental and descriptive approach, system requirements were mapped through structural interviews, direct observation, and targeted surveys with institutional staff. The application features a 5-point biometric face enrollment layout, real-time live-scanner recognition, an 11:59 PM fail-safe cutoff execution script, and a comprehensive administrative portal equipped with leave management, system audit trails, and data filtering capabilities. Empirical validation demonstrates a 90% facial recognition accuracy baseline and a sub-second processing performance profile under indoor ambient light conditions. The proposed system provides a low-cost, low-latency, and internet-independent alternative to manual mechanisms and cloud-dependent services, significantly optimizing administrative transparency and record security within resource-constrained institutional environments.