how to use of STM32cube AI in STM32cubeIDE for STM32H7s78-DK hardware

[Himanshuparmar]

Hello,

I have STM32H7S78-DK board and its have ARM cortex M7 core. As per research I saw that AI model is work on the same M7 cortex but I didn't know how it will work? would you please let us know STM32H7S78 supports AI model implements or not and if yes then can anyone provide steps to how we can implements on the board.

Thanks.

[hamitiya]

Hello,

Yes, STM32H7S78-DK is supported. However, since it is a "flash-less" board, you need to:

- Enable XSPI1 and XSPI2

- Enable EXTMEM_MANAGER and EXTMEM_LOADER

- Prepare memory regions and external RAM / Flash for your target

- For external RAM, add corresponding BSP

These steps should allow you to execute your project in place ( "XIP" ). You can find a project template in STM32H7RS F/W Package example.

It is done automatically in ST Edge AI Developer Cloud if you perform a benchmark or if you generate a project from your own model, and it is also done in X-CUBE-AI when you execute "Validation on Target".

Best regards

[Himanshuparmar]

Hello hamitiya,

thanks for the fast replay.

I need to ask that can we run FP-AI-VISION SDK on STM32H7S78-DK using the changes you have mentioned ? If yes then can you please provide how it will achieve.

[hamitiya]

Hello,

Actually FP-AI-VISION function pack implementation only supports STM32H747I-DISCO. I would recommend to stay with STM32H747I-DISCO if your goal is to run FP-AI-VISION.

Best regards

[Himanshuparmar]

Hello @hamitiya ,

Okay Understood. But we want to make applications like Face Recognition, Face Detection etc. on STM32H7S78-DK. Is it possible to make this kind of applications in STM32H7S78-dk? If no then we have move to STM32H747I-DISCO rather than work on STM32H7S78-DK?

Thanks & Regards

[hamitiya]

Hello,

You have the ability to plug a camera in STM32H7S78-DK  so you should be able to make applications like Face Recognition.

You could go with your own application code on STM32H7S78-DK and get inspiration of the function pack (it is not officially supported, but you can give it a try), or you can use FP-AI-VISION + STM32H747I-DISCO directly. 

Best regards

[Himanshuparmar]

Hello @hamitiya  ,

As per your suggestions, we have picked one AI model from ST Edge AI Developer Cloud and get code for the STM32H7S78-DK. We got some files which is shows how that model uses memory of device. There is no any camera and display interface code in downloaded code from ST Edge AI Developer Cloud. Can you please guide us that how we can send captured image to AI model and got output on display. There is Flower detection model in ST Model Zoo, I have used this model in ST Edge AI Developer Cloud and got STM32CubeIDE project, I have run that project in hardware but in serial terminal only I got this output(please refer this document > stm32h7_log.txt). I have gone through code but there is no any interface of camera and display. Can you please guide us that how we can do this?

Model : fdmobilenet_0.25_128_tfs_int8_image_classification_flowers.tflite

[hamitiya]

Hello,

Indeed, templates from ST Edge AI Developer Cloud do not contain any code related to camera interface. It provides ready-to-try templates to run inference on the target based on model, parameters, and memory configuration of a given target.

Based on your current progression, I will suggest to implement in this template the code associated to data acquisition available in FP-AI-VISION for STM32H747I-DISCO and modify the way you run inference.

Hope it helps.

Best regards,

Yanis

[ishanpadaliya]

Hello Hamitiya,

we are using STM32H7S78-DK. 
we have successfully run the image classification model. But for Object detection model, need to extend the RAM, (need to configure the PSRAM).

How can i configure "Hexadeca‑SPI PSRAM" and extend the RAM. ?
Is it required to use the PSRAM in boot ? ( boot / application )
How to configure the PSRAM in memory mapped mode ?
Any changes required in Tools > Memory Management ?

Best Regards,
Ishan

[hamitiya]

Hello,

To enable PSRAM, you can toggle it in X-CUBE-AI directly (see page 20 here: Getting started with X-CUBE-AI Expansion Package for Artificial Intelligence (AI) - User manual). X-CUBE-AI will copy the BSP and add the init. routine. For your information, it'll also enable memory protection unit region and change access settings in order to increase performance.

On the other hand, you can also see the templates in STM32H7S78-DK driver which uses External PSRAM.

Best regards,

Yanis