Seamless authentication and inspection using facial recognition. This app provides:

• 4 camera feed slots with a dropdown per feed to select the camera (Camo Studio, Iriun, built-in, etc.).
• Live view with live analysis: faces are detected and outlined with a colored perimeter — green = authorized (known face in DB), red = unknown.
• Register face: upload a photo + name to add someone to the known-face database so they show as authorized on the feeds.

• Python 3.10+ and Node 18+ on your laptop.
• Chrome (for localhost and camera access).
• Camo Studio (and/or Iriun): pair your phones so they appear as virtual webcams. The app will list them in the dropdowns once you allow camera permission.

cd backend
python -m venv venv
venv\Scripts\activate   # Windows
# source venv/bin/activate  # Mac/Linux
pip install -r requirements.txt
uvicorn app.main:app --reload --host 127.0.0.1 --port 8000

Leave this running. The first request that triggers face recognition may take a few seconds (model load). The first time you register or recognize a face, the ArcFace ONNX model (~250MB) is downloaded once to backend/data/models/; no C++ Build Tools or InsightFace build is required.

In a second terminal:

cd frontend
npm install
npm run dev

Open http://localhost:5173 in Chrome. Allow camera access when prompted.

1. Register faces: Use the form at the top: enter a name, choose a clear face photo, click “Register face.” Those people will appear as authorized (green) when seen on any feed.
2. Feeds: For each of the 4 slots, open the dropdown and select a camera (e.g. your two Camo cameras). Unused slots can stay on “No camera.”
3. Perimeters: Green box = recognized and authorized; red box = unknown (not in DB or not allowed).

When you add more phones (e.g. 2 more via Camo), they will show up in the same dropdowns; no code changes needed.

• backend/ — FastAPI app: /api/register, /api/recognize, known-face storage under backend/data/.
• frontend/ — React + Vite: 4 feeds, camera dropdown, live view, overlay drawing.
• door-detection/ — Cloned repo: YOLOv8 Door detection (visually impaired). Use for door detection on the camera that faces the door; train once to get doors.pt (see docs/DOOR_MOVEMENT_AND_ACCESS.md).
• docs/DOOR_MOVEMENT_AND_ACCESS.md — Door movement + access: one camera on the person (face + role), one on the door; when the door moves, apply role vs door restrictions and alert. Includes which repo to use and how to detect door movement.
• docs/ZONE_CAMERA_MAPPING.md — How to map floor-plan zones to the camera view (homography approach; optional).
• RECOMMENDATIONS.md — Repo recommendations and what you need to provide for further development.

Use as needed for your hackathon / project.