STEVAL-STWINKT1 ML ON Microprocessor not ISM330DHCLX

[NRaha.1]

I have a STEVAL-STWINKT1 board and I want to create a ML algorithm for the board to conduct predictive maintenance. However, I want my ML algorithm to run on the microprocessor and not on the ISM330DHCLX sensor. This is because, I realized that the ISM330DHCLX sensor is not sensitive to small changes in acceleration (around 0.1g), whereas the motor I am using is very small. Thus I want to use the IIS2DH sensor values as the input to my model instead.

My current workflow is I generate the ML UCF File using Unico, and then use the data logging function pack to flash the model to the STEVAL-STWINKT1 board and monitor its output. However, this workflow does not work if I want my ML algorithm to run on the STM32L4+ MCU itself instead of in the ISM330DHCX sensor.

Is there a step by step example on how I can do this?

[TSand.1]

Hi, you might consider to run the neural network on the STM32 using the X-Cube-AI tool, from which you have to set all the Sensortile.box project. or you might base on the FP-AI-SENSING1 package for already available examples.

https://www.st.com/en/embedded-software/x-cube-ai.html

https://www.st.com/en/embedded-software/fp-ai-sensing1.html

Or alternatively you might use this tool (https://qeexo.com/) that supports the STEVAL-STWINKT1 board and with which you can easily acquire your data, train the NN and downloading the binary file for your tool.