Nshogoza/Face_locking

A research-grade, CPU-optimized face recognition and tracking system implementing ArcFace embeddings, 5-point facial alignment, real-time face locking, action detection, and persistent behavior logging — designed for clarity, performance, and deployment on low-resource machines. Face Recognition & Locking System ArcFace ONNX + 5-Point Alignment + Real-Time Behavior Tracking Author: URUSARO NARASHA Class: Year 3 A Institution: Rwanda Coding Academy A research-grade, CPU-optimized face recognition and tracking system implementing ArcFace embeddings, 5-point facial alignment, real-time face locking, action detection, and persistent behavior logging — designed for clarity, performance, and deployment on low-resource machines. Built for students, researchers, and developers who want serious computer vision without needing a GPU. 📚 Table of Contents Project Overview Core Features Face Locking System System Requirements Installation Guide Project Structure Quick Start Usage Guide Pipeline Architecture Troubleshooting Applications Credits & References 📌 Project Overview This project implements a complete real-time face recognition pipeline using: Haar cascade face detection MediaPipe 5-point facial landmarks Similarity transform alignment (112×112) ArcFace ONNX deep embedding extraction Cosine similarity matching Real-time face locking & behavior analysis It is designed to run entirely on CPU, enabling deployment on standard laptops, making it ideal for education, research, and real-world testing. ✨ Core Features 🎯 Face Recognition Engine CPU-only inference (no GPU needed) 5-point facial landmark alignment ArcFace 512-D embeddings Real-time multi-face detection Open-set recognition (reject unknown faces) Threshold tuning via FAR / FRR evaluation Persistent database storage 🔒 Face Locking & Behavior Tracking (Advanced) Identity-based face locking Real-time head movement tracking Blink detection Smile / expression detection Face distance tracking Automatic behavioral history logging State machine tracking: SEARCHING → LOCKED → LOST 🔐 Face Locking System Once a known identity appears: Face is recognized System locks onto the identity Tracks: Eye blinks Head motion Smiles Face movement Logs timestamped …