platedet

platedet is a lightweight package for detecting license plates in images using an ONNX model. It is designed to be part of a pipeline for detecting, cropping, and reading license plates. The underlying model is just YOLO, converted to ONNX format. The training data comprises primarily Brazilian license plates, sourced from internet images.

Installation

To install the required dependencies, use the following command:

For cpu

pip install "platedet[cpu]"

For cuda 11.X

pip install "platedet[gpu]"

For cuda 12.X

pip install "platedet[gpu]" --extra-index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/onnxruntime-cuda-12/pypi/simple/

Usage

Command Line Interface

You can use the command line interface to detect license plates in an image:

platedet image_path [--model_path MODEL_PATH] [--return_types RETURN_TYPES] [--providers PROVIDERS] [--save_images] [--show_result]

Arguments

• image_path: Path to the input image.
• --model_path: Path to the ONNX model (default: artifacts/platedet.onnx).
• --return_types: Output formats (choices: raw, pil, np_array, boxes).
• --providers: ONNX Runtime providers (default: CPUExecutionProvider).
• --save_images: Save cropped images of detected plates.
• --show_result: Display results with bounding boxes and confidence scores.

Example

To detect license plates and save the cropped images:

platedet path/to/image.png --save_images

To display the results with bounding boxes:

platedet detect.py path/to/image.png --show_result

Using in Code

from PIL import Image
from platedet import Platedet

platedet = Platedet()
image = Image.open('examples/1.jpg')
crops = platedet.inference(image, return_types=['pil'])
for idx, crop in enumerate(crops['pil']['images']):
    crop.save(f'{idx}.jpg')

If you want to use CUDA:

from PIL import Image
from platedet import Platedet

platedet = Platedet(providers=["CUDAExecutionProvider"])
image = Image.open('examples/1.jpg')
crops = platedet.inference(image, return_types=['pil'])
for idx, crop in enumerate(crops['pil']['images']):
    crop.save(f'{idx}.jpg')

Check all execution providers here.