ONNX Runtime Importation (machine learning)

[Dekel]

Hi all.

I'm working on a Visual Studio project, written in C++, which I'd like to import into Segger Embeeded Studio.

I've done this before successfully, but this time the VS project uses a ML library named ONNX Runtime, for which I couldn't find an embedded counterpart.

Does anyone have an experience with importing a ML project into Segger? Specifically ONNX Runtime?

Alternatively, I've found this guide for importing an ONNX model into TensorFlow. Is importing a TensorFlow project into Segger possible?

What are my options?

Thanks,

Dekel

[Øyvind]

Hello Dekel, 

Have you looked at our Machine Learning solution Edge Impulse, which is a leading development platform for machine learning on edge devices i.e. embedded. Read more about How TensorFlow helps Edge Impulse make ML accessible to embedded engineers, and our blog post Machine Learning with the Thingy:53.

What are my options?

We offer two samples and a full application to run on several of our development kits.

• Edge Impulse: Data forwarder
• Edge Impulse: Wrapper

• nRF Machine Learning

Kind regards,
Øyvind

[Dekel]

Thanks a lot for your detailed reply!

As it seems to me, Edge Impulse is only relevant to nRF Connect SDK, and not to nRF5 SDK V16, which is the one I'm working with.

Did I get this right?

[Øyvind]

We do not have machine learning support in the nRF5 SDK. 

Kind regards,
Øyvind