You can deploy this model and start earning money today!

larshiakzemil/
paddle_rb_scanner_yolo_plate

Name: paddle_rb_scanner_yolo_plate
Availability: InStock

onnx

Views

No views yet

Logo

Multi-Camera License Plate Recognition

FastAPI service for synchronized multi-camera license plate detection + OCR, with MinIO image upload and WebSocket result streaming.

What It Does

Reads multiple RTSP/video sources from configs/config.yaml
Synchronizes frames across cameras (src/frame_gather.py)
Detects plates with YOLO (src/detector_yolo.py)
Recognizes text with ONNX RobustScanner (src/recognizer.py)
Filters detections by confidence (0.94 in src/aysnc_server.py)
Draws boxes/text and uploads annotated frames to MinIO
Streams result payloads over WebSocket (MsgPack)
Serves API docs and a static dashboard page

Repository Layout

plate_server.py: FastAPI app + API endpoints
src/aysnc_server.py: async processing service (LicensePlateServer)
src/frame_gather.py: synchronized frame capture
src/run_yolo.py: detector + recognizer pipeline
src/detector_yolo.py: Ultralytics YOLO wrapper
src/recognizer.py: OCR pipeline
src/minio_uplaoder.py: MinIO clients
configs/config.yaml: runtime config
dashboard/index.html: static dashboard
docker-compose.yml: app + MinIO

Requirements

Python 3.10+
Linux/macOS recommended for OpenCV/video tooling
Model files present:
- models/yolo12n_openvino_model/
- models/rb_scaner.onnx

Local Run

Install dependencies:

bash

python -m venv .venv
source .venv/bin/activate
pip install --upgrade pip
pip install torch torchvision --index-url https://download.pytorch.org/whl/cpu
pip install openvino msgpack
pip install -r requirements.txt

Start API server:

python plate_server.py

Server runs on 0.0.0.0:8003.

Docker Run

docker compose up --build

Services:

API: http://localhost:8003
Swagger UI: http://localhost:8003/docs
Dashboard page: http://localhost:8003/dashboard
MinIO API: http://localhost:9000
MinIO console: http://localhost:9001

Configuration

Main config file: configs/config.yaml.

Current expected fields include:

yaml

rtsp_urls:
  - id: 11
    url: /path/to/source1.mp4
  - id: 12
    url: rtsp://user:pass@camera/stream

log_level: INFO
endpoint: 0.0.0.0:9003
access_key: minioadmin
secret_key: minioadmin
bucket: detector-frames
dashboard_ws: ws://<dashboard-host>:8075
upload_interval: 5

Notes:

rtsp_urls is required for frame ingestion.
The service currently initializes MinIO with hardcoded endpoint/credentials in src/aysnc_server.py (0.0.0.0:9000, minioadmin/minioadmin) and uses bucket from YAML.

API

`POST /config`

Create/replace config file (JSON or YAML body). Refuses update while service is running.

`PUT /config`

Alias of POST /config in current implementation.

`POST /start`

Start frame sync + processing + websocket server.

Body (UploaderConfig):

json

{
  "upload_interval": 5.0,
  "download_verify": false,
  "download_dir": "downloaded",
  "target_width": 640
}

`POST /stop`

Stop background processing.

`GET /health`

Health + running flag.

`GET /get-config`

Returns loaded YAML config + runtime config.

WebSocket Streams

Internal result stream: ws://<host>:9101
- Broadcast from LicensePlateServer
- Payload format: MsgPack
- Example top-level shape:

json

{
  "results": [
    {
      "cam_id": 11,
      "ocr": "12A34567",
      "confidence": 0.97,
      "minio_path": "cam_11/2026-02-22/1740....jpg",
      "bbox": [100, 200, 260, 260]
    }
  ]
}

External dashboard forward target: dashboard_ws from config
- Same packed payload is forwarded when available.

Quick API Examples

Write config:

bash

curl -X POST http://localhost:8003/config \
  -H "Content-Type: application/json" \
  -d '{
    "rtsp_urls": [
      { "id": 11, "url": "/home/arshia/Downloads/parking-1.mp4" },
      { "id": 12, "url": "/home/arshia/Downloads/video.mp4" }
    ],
    "log_level": "INFO",
    "endpoint": "0.0.0.0:9003",
    "access_key": "minioadmin",
    "secret_key": "minioadmin",
    "bucket": "detector-frames",
    "dashboard_ws": "ws://192.168.7.119:8075",
    "upload_interval": 5
  }'

Start processing:

bash

curl -X POST http://localhost:8003/start \
  -H "Content-Type: application/json" \
  -d '{"upload_interval": 5.0, "download_verify": false}'

Stop processing:

curl -X POST http://localhost:8003/stop

Output and Artifacts

Cropped plates: cropped_plates/
Optional verification downloads: downloaded/ (when enabled)
Optional synchronized tile snapshots: path from output_dir in config

Known Caveats

Filename typo in source is intentional in current repo: src/aysnc_server.py.
MinIO env vars in docker-compose.yml use MINIO_ENDPOINT=minio:9001, but MinIO object API is usually 9000.
Swagger endpoint is custom-mounted at /docs.

Troubleshooting

Service won’t start:
- Ensure model files exist in models/
- Check camera/video paths in configs/config.yaml
- Verify MinIO is reachable and bucket is valid
No detections:
- Lower detection threshold in detector flow if needed
- Confirm input frames are non-black and readable
No dashboard updates:
- Confirm dashboard_ws is reachable
- Check WebSocket clients are connected to :9101 if using internal stream

larshiakzemil/paddle_rb_scanner_yolo_plate

Multi-Camera License Plate Recognition

What It Does

Repository Layout

Requirements

Local Run

Docker Run

Configuration

API

POST /config

PUT /config

POST /start

POST /stop

GET /health

GET /get-config

WebSocket Streams

Quick API Examples

Output and Artifacts

Known Caveats

Troubleshooting

larshiakzemil/
paddle_rb_scanner_yolo_plate

`POST /config`

`PUT /config`

`POST /start`

`POST /stop`

`GET /health`

`GET /get-config`