Object detection: deploy custom tiny ISSIMO-V8 model on STM32H747I-DISCO board

[syrine]

Hi Community,

I’m working on deploying my custom tiny ISSIMO-V8 model for object detection onto the STM32H747I-DISCO board with the OMV CAMS camera. Can I use the FP-AI-Vision 1 project, which is based on object detection for people counting, and modify it to adapt to my model (detecting kitchen utensils and multiple objects)?

Thank you in advance for your guidance!

Best regards,

[Julian E.]

Hello @syrine,

You should look at ST Model Zoo for object detection: https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/object_detection 

And in particular, the deployment operation mode that allow you to deploy your model on the  STM32H747I-DISCO that you happen to use: https://github.com/STMicroelectronics/stm32ai-modelzoo/blob/main/object_detection/deployment/README.md 

When using the deployment operation mode and scripts, it will modify the application project located here: https://github.com/STMicroelectronics/stm32ai-modelzoo/tree/main/stm32ai_application_code/object_detection 

So you can have access to the source code (the deployment scripts directly flash your board)

Have a good day,

Julian

[syrine]

Thank you for the response. Is there no possibility to start the project from scratch without using the Model Zoo, and instead begin from the generated project using X cube Ai ?

[Julian E.]

Hello @syrine,

You can start the project from scratch using cubeMX and X cube AI if you want.

You can take a look here: https://wiki.st.com/stm32mcu/wiki/AI:How_to_measure_machine_learning_model_power_consumption_with_STM32Cube.AI_generated_application 

Have a good day,

Julian