syanng/onnx-engine-boilerplate

Boilerplate to package ONNX model with pip

ONNX Engine Package

This is a boilerplate for PIP Package Model ONNX. The model will simply load onnx package and return ReadyOCR python object.

If you want to go deeper, review this blog post from RealPython.

Edit Package

Go to the file src/onnxengine/__init__.py and edit ONNXEngine class. Please remember to update version in pyproject.toml for each update.

Install Locally

After git clone, you can access the codebase and simply run the following command line:

python -m pip install -e .

If you want to add dependancies, go to file pyproject.toml and edit lines:

dependencies = [
    "onnxruntime",
    "readyocr"
]

Basic Usage

After installing, we can use the ONNX engine as following:

from onnxengine import ONNXEngine

model = ONNXEngine(model_path="any_model.onnx")
model.run(image_path)

Or you can run examples/sample.py file as following:

python sample.py -m `link/to/model.onnx' -i 'link/to/image'