Overview of ONNX

For those who have never used AI in ONNX format, here is a brief overview of ONNX.
If you know the basics of ONNX, you don't need to read this.

What's ONNX?

ONNX
= Open Neural Network eXchange

• A format to use AI in various environment.
• You can create files in this format from various machine learning frameworks, and
  you can run AI on various platforms with this format.

(https://onnxruntime.ai/docs/execution-providers/ より引用)

How to get ONNX file

You can make it yourself from PyTorch, TensorFlow, etc. Also there are kind people who publish ONNX format AI for free.

• (Example 1) ONNX Model Zoo: ONNX official
• (Example 2) PINTO model zoo: A repository where PINTO shares the results of optimization of various models. Various formats are provided, including ONNX.

Check the license carefully before using them.🙏

How to use ONNX

Roughly there are 4 steps.

1. Load the ONNX file.
2. Specify the input data source.
3. Specify the output data destination.
4. Run.

See the Official Documentation for the details.
NNEngine provides convenient functions to execute 1 to 4 above.