“Watchwise: A Real-Time Face Recognition and Anti-Spoofing System for Secure Person Tracking and Identity Logging”

The rise in demand for secure, automated, and contactless identity verification systems has driven the need for intelligent biometric solutions. This paper presents Watchwise, a real-time face recognition system designed for person tracking and identity logging using advanced deep learning and computer vision techniques. The system employs MTCNN (Multi-task Cascaded Convolutional Networks) for accurate face detection and alignment, followed by FaceNet for extracting robust $\mathbf{1 2 8}$-dimensional facial embeddings. Identity matching is performed using cosine similarity against stored embeddings to verify individuals efficiently. To prevent unauthorized access and spoofing attempts, the system integrates an anti-spoofing module comprising blink detection, texture analysis using CNNs, and optional infrared sensing. A lightweight SQLite database is used for secure local storage of embeddings and activity logs. The system is benchmarked using a combination of standard datasets such as LFW and CelebA, along with custom datasets to evaluate performance in practical conditions. Metrics including accuracy ($95.2 \%$), precision, recall, F1-score, FAR, and FRR confirm the system’s effectiveness. Furthermore, the solution is optimized for real-time operation on edge devices such as Raspberry Pi, making it suitable for resource-constrained environments. The system presents a scalable, privacy-aware solution for secure person tracking applications in educational, corporate, and security domains.