Is it possible to use multiple AI models on my STM32N6570-DK?

[lyannen]

Hello,

im trying to implement an ai project that uses two different ai models in one pipeline.
The output from model A should be the input for model B.

The question that im asking would be how im able to combine the two models within one project in my STM32CubeIDE in order to implement such pipeline where A is the input of B. Or rather the output from A triggers B.

Any help is appreciated!

Kind Regards!

[Julian E.]

Hello @lyannen,

I would suggest looking at this getting started:

GitHub - STMicroelectronics/x-cube-n6-ai-hand-landmarks: An AI software application package demonstrating hand landmarks detection use case on STM32N6 product.

In this, 2 models are used, the first one to detect a hand in the image and the second one uses the detected hand as input to then be put landmarks on it. 

It thinks, there are processing in between the 2 models to replace the hand before giving it to the second model.

Have a good day,

Julian

[lyannen]

Thank you very much for the quick response Julian, this is a great starting point for me!

I hope that this answers most of my questions. However this project has no .ioc file which makes it a little harder for me to set up a project from scratch. But i already managed to add multiple models with X-CUBE-AI in the .ioc so ill hopefully manage. 
If not i will let you know any further questions.

Kind regards!

[Julian E.]

@lyannen, 

Indeed, the ioc is not present (I never remember the reason, but there is a good one).

We will very soon publish a tutorial starting in CubeMX from scratch and creating an application on the Nucleo and DK board. The application and model weights are stored in external flash and loaded by a FSBL.

I will update you when the tutorial is published.

With this and the getting started, you should have everything you need to hopefully create your application.

Have a good day,

Julian

[lyannen]

Hello Julian,

any Update on when the tutorial will be published?

Kind regards!

[Julian E.]

Hello @lyannen,

As of today, it is a draft waiting for validation.

Have a good day,

Julian

[PeleAB]

I have done so in my project and created an automated pipeline that lets you add models paths to json file and generate multi model application: https://github.com/PeleAB/STM32N6-FaceRecognition

[lyannen]

Hi @PeleAB,

what exactly do you mean by " lets you add models paths to json file and generate multi model application"?
Does that mean that i can simply exchange the models in the repo to different models (even different use cases) and run the project?

For now your repo is a great start for my own implementation that varies a lot from your implementation however.

Kind regards!

[Julian E.]

Hello @lyannen,

You can find the tutorial I mentioned here.

Sorry, it took more time than expected.

Nucleo N6:  How to build an AI application from scratch on the... - STMicroelectronics Community

DK N6: How to build an AI application from scratch on the... - STMicroelectronics Community

Have a good day,

Julian