does stm32 model zoo support board other than boards specified below?

[NewSTM32Gamer]

D:\stm32ai-modelzoo\object_detection\scripts\deployment> python .\deploy.py

[INFO] : Successfully connected, starting the model validation on target ...
[INFO] : Starting the benchmark on target STM32H747I-DISCO, other available boards ['STM32F469I-DISCO', 'STM32L4R9I-DISCO', 'STM32H573I-DK', 'STM32H7B3I-DK', 'NUCLEO-F401RE', 'NUCLEO-G474RE', 'B-U585I-IOT02A', 'NUCLEO-H743ZI2', 'STM32F769I-DISCO', 'STM32H735G-I2', 'STM32F769I-DISCO', 'STM32H735G-DK', 'STM32H747I-DISCO', 'STM32F746G-DISCO']

I want to train and deploy an object detection model on my new purchased NUCLEO-H745ZI-Q board. I am reading the code line by line to see if I can adjust the code to train and deploy the model without getting into the STMCubeIDE.
And I see there are other boards available for benchmarking, I am wondering if I can read model zoo repo line by line and see if I can change the code so that it could train and deploy object detection or image classification models on NUCLEO-H745ZI-Q board  using customized dataset without actually entering CubeIDE? I am worried that if there exists some hidden lines of code that will prevent boards that are not specified from deployment. 

I understand I may need configure the board in the CubeIDE first to generate files needed by the NUCLEO-H745ZI-Q board, but once this is done, I should be able to train, evaluate and deploy the model using command lines easily.

Thanks in advance for your answer, Hope to hear from you soon! If there is any confusion in my question, please do not hesitate to speak out! 

[NewSTM32Gamer]

The repository I mentioned above is: stm32ai-modelzoo
https://github.com/STMicroelectronics/stm32ai-modelzoo

[Matthieu]

Hello,

The model zoo allows to deploy a full application, including data acquisition, pre-processing, display.

So the deployment is focused on configurations that can handle these end-to-end use cases.

For instance, for image classification, the STM32H747I-DISCO with B-CAMS-OMV camera daughter board or the NUCLEO-H743ZI2 with B-CAMS-OMV camera daughter board and USB display output.
You will find the list on the "available use-cases section of the readme.

But using the STM32Cube.AI Developer Cloud https://stm32ai.st.com/stm32-cube-ai-dc/ allows to generate the C code for any STM32. You can also generate a template project.

If you prefer to use Python API, you can refer to this Jupyter notebook: https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/tutorials/notebooks/STM32CubeAI_DevCloud.ipynb

Best Regards

 

[NewSTM32Gamer]

Thank you so much for your prompt reply!

I tried STM32Cube.AI Developer Cloud and it works out as a template project. My goal is to to deploy a full application, including data acquisition, pre-processing, display as you have mentioned.  And I would like the training and deployment to be focused on configurations that can handle these end-to-end use cases. Is there any further instructions on how this goal can be achieved? The STM32 model zoo does not support all boards and I want to use my newly purchased boards to achieve the same end to end use cases just like the STM32H747I-DISCO with B-CAMS-OMV camera daughter board example using the model zoo. 

I know this is gonna to be a difficult approach to deploy the model, but Thanks for any clarifications in advance!

[Matthieu]

Hello,

I would suggest starting from the model zoo Getting Started package and port to other boards.

For computer vision, we especially have one example on NUCLEO-H743ZI2 with B-CAMS-OMV camera daughter board and USB display output. It is close from your NUCLEO-H745ZI-Q board.

Best Regards,