Install

anpr-pipeline needs Python 3.11 to 3.14. The models download on first use (about 30 MB) and are cached after that.

From PyPI

pip install anpr-pipeline

This installs the anpr command. Check it:

anpr version

Optional extras

Extra	What it adds
anpr-pipeline[gpu]	onnxruntime-gpu, for running the models on an NVIDIA GPU.
anpr-pipeline[otel]	OpenTelemetry tracing for the API. See Monitoring.

With Docker

The repository has a Dockerfile and a docker-compose.yml. The service needs a secret to hash plates with, called the pepper. Generate one and start the service:

git clone https://github.com/mftnakrsu/Automatic_Number_Plate_Recognition_YOLO_OCR.git
cd Automatic_Number_Plate_Recognition_YOLO_OCR
export ANPR_PLATE_HMAC_PEPPER="$(python -c 'import secrets; print(secrets.token_hex(32))')"
docker compose up --build

Open http://localhost:8000. The database lives in the anpr-data volume, so it survives restarts. Keep the pepper: with a different one, the same plate gets a different hash. See Privacy for why.

From source

You need uv.

git clone https://github.com/mftnakrsu/Automatic_Number_Plate_Recognition_YOLO_OCR.git
cd Automatic_Number_Plate_Recognition_YOLO_OCR
uv sync --all-extras --group dev
cp .env.example .env
uv run anpr generate-pepper   # paste the output into .env as ANPR_PLATE_HMAC_PEPPER
uv run anpr serve

Next: read your first plate.