How to run C/C++ model created using Edge-impulse on Micropython

[ujur007]

I am here to ask one question. For my current project, I am using ESP32 and micropython. I am quite enjoying it. But now I have a trained ML model from Edge Impulse that I would like to deploy to this device and run the live classification. The issue is that the sources of the trained model might be so much and also there may be some assembly code in it.

I just want to find out whether it is possible to run such models using Micropython?

https://docs.micropython.org/en/latest/develop/cmodules.html

[jonnor]

I do not think there are any projects that provide Edge Impulse support for MicroPython. But it is theoretically possible to create MicroPython modules by wrapping the C++ code that they provide. The concerns are similar as to that of https://github.com/orgs/micropython/discussions/18273

[ujur007]

@jonnor Thanks, possibly I will try that. At the moment I am using C++ SDK provided by Edge Impulse.
I tried Emlearn micropython and microphone sample and that works fine but there is just DSP and no inference.
Also, there was some memory related error in that sample to run it continually.

[jonnor]

@ujur007 nice that you got the examples running! Your goal/task is to run some audio classification or other ML on microphone signal? There is an open issue for that in emlearn-micropython, see emlearn/emlearn-micropython#6
So it is planned, but I cannot give a specific timeline for it