Face Recognition System for Campus

This report presents a design for an attendance system using device cameras (student laptops or smartphones) and optional fixed cameras. The system captures faces, detects them, and identifies each person via deep-learning embeddings (e.g. FaceNet) against an enrolled database. Liveness checks (texture analysis, eye-blink detection, etc.) ensure security against photo/video spoofing. Attendance entries (timestamps) are logged in a database. Only authenticated teachers or admins can access the system. The implementation uses open-source tools (OpenCV, TensorFlow/PyTorch, Flask) and low-cost hardware (Android/iOS devices, Raspberry Pi, Jetson Nano). Key evaluation metrics include recognition accuracy, FAR/FRR, latency, and throughput. Based on similar systems, we expect >95% accuracy and sub-second processing. Privacy is handled by storing only encrypted embeddings and following India's DPDP 2023 guidelines. We include architecture diagrams, comparative tables (algorithms/datasets/hardware), and a development timeline. We propose an automated attendance system leveraging students' own devices (laptop/phone cameras) and optionally campus cameras (CCTV/RTSP). Attendance is recorded twice daily without manual roll calls. The software runs on personal devices (browser or app) for convenience. Using lightweight CNN models (MobileNet, FaceNet, etc.), the system achieves high accuracy. Anti-spoofing (texture CNN + blink/motion checks) is integrated to prevent fraudulent sign-ins. A secure portal allows only teachers/admins to log in and view records. We detail system requirements, methodology, architecture, implementation, evaluation plan, and ethical considerations. Data privacy is a priority, with encrypted face embeddings and DPDP 2023 compliance (explicit consent, data minimization). The design is scalable to large classes and multi-camera setups (including existing CCTV).