itsrishu02/cctv-person-detection-api

ZEEX AI Face Recognition v0

This is the Vercel/v0-style version of zeex-face-recognition.

It keeps the YOLO/OpenCV recognition pipeline in a local FastAPI backend and replaces the Streamlit UI with a Next.js frontend that can be edited in v0 or deployed on Vercel.

What changed

• —Empty frames are now reported as EMPTY_ZONE.
• —The annotated frame shows EMPTY ZONE / NO PERSON when no face and no person detection exists in the observed frame.
• —Empty-zone detection has a tunable Empty Zone Frames threshold in the UI.
• —Streamlit was removed from this version.
• —The UI lives in frontend/ as a Next.js app.
• —The recognition loop lives in backend/api.py as HTTP endpoints.

Project layout

zeex-face-recognition-v0/
  backend/
    api.py
    face_pipeline.py
    recognizer.py
    encode_faces.py
    config.yaml
    encodings.pkl
    models/
    requirements.txt
  frontend/
    app/
    package.json
    .env.example

Run locally

Open one terminal for the backend:

cd C:\Users\RishuSingh\Downloads\Mins_project\zeex-face-recognition-v0\backend
python -m pip install -r requirements.txt
python -m uvicorn api:app --host 127.0.0.1 --port 8000 --reload

Open another terminal for the frontend:

cd C:\Users\RishuSingh\Downloads\Mins_project\zeex-face-recognition-v0\frontend
npm install
npm run dev

Then open:

http://localhost:3000

Choosing a source

Changing Camera ID in the frontend automatically switches to the matching RTSP source and starts recognition:

CAM-01 -> rtsp://65.1.214.31:8554/cam1
CAM-02 -> rtsp://65.1.214.31:8554/cam2
CAM-10 -> rtsp://65.1.214.31:8554/cam10

The log line below means the backend is working, but the RTSP endpoint itself is not serving a stream at that path:

method DESCRIBE failed: 404 Not Found

Use the Path tab with the configured sample video, use Upload, or replace the RTSP URL with one that VLC can open successfully. Common RTSP paths look like /Streaming/Channels/101, /cam/realmonitor?channel=1&subtype=0, or a camera-specific path from the device settings.

Empty-zone threshold

Empty Zone Frames controls how many consecutive processed frames must contain zero detected people before the UI/API reports EMPTY_ZONE. Increase it if empty-zone status flickers during brief detector misses; decrease it if you want faster empty-zone detection.

Backend API

• —GET /api/health
• —GET /api/defaults
• —GET /api/workers
• —POST /api/videos/upload
• —POST /api/recognition/start
• —POST /api/recognition/stop
• —POST /api/recognition/settings
• —GET /api/recognition/status
• —GET /api/recognition/frame

Vercel note

Deploy frontend/ to Vercel and set:

NEXT_PUBLIC_API_BASE_URL=<your backend URL>

The Python backend should run on a machine/server that can access the RTSP camera, local video files, model files, and OpenCV dependencies.