"Correct" Way to Gather Data and Train ML for IMU

[colin_catlin]

I am lost in the maze of documentation for AI features at STM. 

I have access right now to:

• STEVAL-MKBOXPRO (SensorTile.box PRO)
• Custom board with LSM6DSV16X

I want to:

• gather IMU data to train the model
• train the IMU device decision tree classifier (3 class)
• deploy it ultimately on the custom board, but the MKBOXPRO is also acceptable as a target for now

My background is in data science, so the 'AI' part I understand. My embedded programming and hardware design knowledge is decent but still beginner level compared to most people here.

So far:

• Unico-GUI - doesn't work on the mkboxpro?
• MEMS studio doesn't have a data collection feature like Unico? And I'm having connection issues, although I suspect I'll figure that out soon enough
• Apps for mkboxpro on phone don't seem to look like they do in the documentation, perhaps due to it not being in Pro mode
• There's yet another tool called STEdge AI Core which I haven't explored yet but seems to overlap with the others

The simplest path I see right now might be to buy a STEVAL-MKI109V3 and use it with Unico instead of the MKBOXPro.

My question is, what is the 'correct' way as of late 2024 to do what I am trying to do. Can I do this with the MKBoxPro?

[SimonePradolini · Accepted solution]

Hello @colin_catlin 

The idea behind Edge AI Suite is to provide a comprehensive solution to help developers accelerate their product transformation. Here, developers can find the tools to optimize and deploy machine learning algorithms, from data collection to final deployment on hardware. 

This presentation can guide you to select the tool you need.

I guess you are searching for the High Speed Datalog tool, that allows users to manage the acquisition and labelling of sensor datasets via SD card or USB for different STEVAL and NUCLEO boards, included SensorTile.box PRO (STEVAL-MKBOXPRO). The firmware is released both in source code or precompiled, so if you are not interested in embedded programming you are free to flash the precompiled binary directly. It offers also a Python SDK to collect and manage your dataset and convert them in the preferred format. Converters to CSV, Unico and NanoEdge format are natively supported.

Let me know it this solution can help you.

Best regards

Simone