Resti Marlin
Sekolah Tinggi Manajemen Informatika dan Komputer Jakarta STI&K

Published : 1 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 1 Documents
Search

Facial Recognition System Based on YOLOv5, FaceNet, and Anti-Spoofing Mechanisms for Room Security Resti Marlin; Hariyanto
JURIKOM (Jurnal Riset Komputer) Vol. 13 No. 5 (2026): Oktober 2026
Publisher : Universitas Budi Darma

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30865/jurikom.v15i5.10151

Abstract

Physical security in restricted areas requires fast, accurate access control capable of preventing presentation attacks. This study develops an intelligent room access control system integrating Deep Learning and anti-spoofing mechanisms. The system employs YOLOv5 for precise face detection, followed by feature extraction using FaceNet and dimensionality reduction via Principal Component Analysis (PCA) to 50 principal components. To prevent spoofing via printed photos or recorded videos, eye-blink detection using MediaPipe Face Mesh and Eye Aspect Ratio (EAR) is implemented. System monitoring and access logs are managed centrally via an SQLite database with automated web server notifications. The dataset comprises 50 recognized employee subjects (12 training and 3 testing photos per subject) alongside unregistered guest samples. Experimental results demonstrate that the system achieves high verification accuracy with an Euclidean distance threshold of 0.75, operating stably within a distance of 50 cm to 150 cm and under lighting conditions between 350 and 450 lux. Furthermore, the system demonstrates high resilience in mitigating photo and video presentation attacks under evaluated configurations. However, operational limitations occur at distances exceeding 200 cm or under low-light/infrared conditions (< 10 lux) with heavy facial occlusions, yielding inconsistent results. In conclusion, the proposed system successfully enhances physical security standards for restricted rooms under controlled operational conditions.