Model Zoo examples compile but fail to run

[Steve6]

Hello,

I am currently testing out the model zoo examples, following https://community.st.com/t5/stm32-mcus/how-to-run-ai-models-from-model-zoo-on-stm32n6/ta-p/814254 and the READMES. Programming the .hex files to the STM32N6-DK works and produces a running firmware. However, when I compile the application using the provided makefile and flash it onto the MCU as described in the README, the screen just remains black. I tested this for object-detection, face-detection and pose-estimation.

Best,

Steve

[rfronteddu]

Which board are you using? I tested the object detection project provided here and also this one and both work with the latest stack provided by ST. To do a quick test you can simply open with cube ide the .project file build and run. Provided that your board is setup correctly you will see the test app running.

Note that for the tutorials, you must use the exact same versions used by them for all components or you may very likely run into trouble.

[Steve6 · Accepted solution]

I am using the STM32N6570-DK. It was the wrong version of the STM32CubeProgrammer. Thank you for the hint. I had the newest version 2.21.0. When signing the binary, one has to add the -align option to make it work. Why is a repo updated three weeks ago dependent on a software that is three versions old???

The second error is that the command that: https://community.st.com/t5/stm32-mcus/how-to-run-ai-models-from-model-zoo-on-stm32n6/ta-p/814254 provided to convert the example .tflite model to code is wrong for the newest version. It is missing the option --output-data int8. The correct command is:

stedgeai generate --model st_yolo_x_nano_480_1.0_0.25_3_st_int8.tflite --target stm32n6 --st-neural-art default@user_neuralart_STM32N6570-DK.json --input-data-type uint8 --output-data-type int8

[rfronteddu]

From what I am observing, I think that the toolchain is evolving fast and they haven't figured out all the kinks yet. 

I was told they are working on some new tutorials so I hope they will use the latest versions for those!

[Julian E.]

Hi,

For each version of the ST Edge AI Core, the Model zoo and the application package are updated at the same time (or shortly after). 

In all github (model zoo or package), the version of the core that is supported is written. 

So please make sure that you are using the correct versions.

There can still be bug, so feel free to report anything that you suspect being wrong.

Have a good day,

Julian