A secure, on-premise, contactless face recognition system designed for university student authentication, ensuring real-time performance (<5s), data privacy, and KVKK/GDPR compliance. Developed as part of a graduation project and presented via poster and thesis presentations.
Real-Time Face Detection & Recognition: Uses OpenCV Haar Cascade for CPU-efficient face detection and FaceNet embeddings for accurate identity verification.
On-Premise Deployment: All components (web server, database, ML models) run locally, ensuring full data control and legal compliance.
Web-Based Dashboard: Role-specific interfaces for Security Personnel, Managers, and Admins. Real-time log monitoring, user management, and reporting.
Secure Authentication: Passwords hashed with bcrypt, Two-Factor Authentication (TOTP via pyotp), and Role-Based Access Control (RBAC).
Efficient Similarity Search: PostgreSQL + pgvector extension to store 512‑dimensional embeddings and perform fast KNN queries (<0.7 threshold).
Asynchronous Updates & Reporting: AJAX (Fetch API) for live log updates; ReportLab for PDF report generation.
Security Personnel: View live logs and video feed at /guard/dashboard.
Manager: Monitor system status and generate CSV/PDF reports at /manager/dashboard.
Admin: Register students, manage users, and load embeddings at /admin.
Response Time: 2.5–3.5s (<5s target)
Accuracy: ≥70% genuine pairs below threshold in 30‑student dataset
User Acceptance: All core scenarios passed (authentication, registration, RBAC)
KVKK Compliance: On‑premise data storage, informed consent, data minimization
Prof. Dr. Mehmet Devrim Akça (Supervisor)
Volunteer students for KVKK‑consented data collection
Team members: Duygu Önder, İsa Berk Geriler