Facial Registration System

A real-time face registration and recognition system built with:

• YuNet (OpenCV) — lightweight face detector with 5-point landmark detection
• SFace (OpenCV) — 128-dimensional face embedding model
• FAISS — fast approximate nearest neighbor search
• CustomTkinter — modern dark-mode GUI

Architecture

Camera → YuNet Detector → alignCrop → SFace Embedder → FAISS Index
                                                              ↓
                                               FaceChecker.search_face()

Setup

See ENVIRONMENT_SETUP.md for environment setup instructions. See MODEL_DOWNLOAD.md to download the required ONNX models.

Running

python registration_app.py  # register students
python check_face.py        # check/verify faces

Evaluation

python tests/evaluate_lfw.py  # run benchmark on Olivetti / LFW

Project Structure

cam_on.py            - Camera, YuNet detector, SFace embedder, FAISS index
check_face.py        - FaceChecker FAISS nearest-neighbor search
registration_app.py  - CustomTkinter registration GUI
logger.py            - dual file+console logging
tests/               - unit tests and evaluation scripts