stm32n6 upgrading GettingStarted-ObjectDetection to ST Edge AI 3.0 runtime error

[svogl]

I have a small problem upgrading the STM32N6-GettingStarted-ObjectDetection example project to a newer Edge AI version: I am trying to upgrade the existing yolo person detection net to the latest Edge AI revision (3.0) before exploring custom network.

I think I updated all necessary parts as described in the Edge AI version update FAQ

• new static library (the non- _PIC version)
• updated the NPU subdirectory, added missing file to the STMCubeIDE project
• in main.c
  • changed the NN name from 'Default' to 'network' to match the generated code
  • replaced npu_cache_init with npu_cache_enable (at this point, the inference loop was stuck waiting for completion , so..: )
  • implemented npu_cache_enable_clocks_and_reset with the original CACHEAXI enabling code

At this point, the npu seemingly started but causes this error:

Bus interface interrupt
 BUSIF0 ERR: 0x0
 BUSIF1 ERR: 0x3
 ATON_STD_IRQHandler()@877: irqs=0xlx
 assertion "0" failed: file "/home/simon/STM32CubeIDE/workspace_1.19.0/STM32N6-GettingStarted-ObjectDetection/Middlewares/AI_Runtime/Npu/ll_aton/ll_aton_runtime.c", line 881, function: __LL_ATON_RT_IrqErr

Obviously I am missing some init function that the new library version needs but that has not been called, any hints are welcome!

My current state can be found in this branch:

https://github.com/svogl/STM32N6-GettingStarted-ObjectDetection/tree/feature/model-update

 

Side-notes:

• I've tried the latest CubeMX but that has a) an older AI code release, b) does not generate AI code for cmake projects
• The board & software is pretty new so much is in flow & converging, I can happily deal with that; if there's a better code base to start I can move my stuff over without problems.

Happy hacking & thanks from Austria

Simon

 

 

[Julian E.]

HI @svogl,

 

The getting started should be upgraded to the core 3.0 on Thursday (22nd). Could you please wait until they are released?

 

The getting started packages exist as standalone but they are also part of the ST Model Zoo services. Normally the github should be updated on the same date, but if it is not the case, I invite you to look at the model zoo repo first:

stm32ai-modelzoo-services/application_code at main · STMicroelectronics/stm32ai-modelzoo-services · GitHub

 

The core 3.0 brings a new API, I don't think the previous version of the getting started packages used it. There are quite a few changes. The new packages are using this new API.

 

Have a good day,

Julian

[svogl]

Ok great :)
I've moved to a vscode / cmake based approach by now and include all depencies as git submodules that should allow for a quick upgrade.

Thanks Simon  

[Julian E.]

Hi @svogl,

 

Did you manage to solve your issue?

 

Have a good day,

Julian

[svogl]

Hi Julian, no this is still open, we tackled other issues meanwhile and will come back to this shortly!
Thanks for checking back,

Simon