I have the STM32N6570-DK board, and I was able to execute the example from:
- https://github.com/STMicroelectronics/STM32N6-GettingStarted-ObjectDetection
Following the steps provided in the workshop:
- https://www.youtube.com/watch?v=29EUWNwfywI&list=PLTJzs51NlEICI7ARAHOyUqMhTgJE0UFv0&index=6
- https://www.youtube.com/watch?v=VV9f5e3nqM4&list=PLTJzs51NlEICI7ARAHOyUqMhTgJE0UFv0&index=7
---
However, after I was able to execute it at first, I tried to convert my own model and modify the example just to execute my own Yolo model, just to execute the inference and perhaps get the inference time via UART.
The conversion worked fine using the stedgeai (version 2.2) command; however, it didn't work on the board on the first try, which was expected.
After a few failed tries, I removed all of my local changes and loaded the initial example binary, the "Model/STM32N6570-DK/network_data.hex" and the application binary again, using the board in development mode. To my surprise, the original example isn't working anymore. I even tried to erase the flash.
Therefore, I removed the call to "Run_Inference()" and "app_postprocess_run()" at these lines:
- https://github.com/STMicroelectronics/STM32N6-GettingStarted-ObjectDetection/blob/main/Application/STM32N6570-DK/Src/main.c#L209-L213
And the example started to work again without the model inference execution.
---
To summarize my steps to flash the ObjectDetection example were:
- Build the original ObjectDetection example;
- Set the board to Development Mode (BOOT0 Low and BOOT1 High);
- Use the STM32CubeProgrammer to flash the network_data.hex;
- Flash the application binary at "./Application/STM32N6570-DK/STM32CubeIDE/Debug/STM32N6570-DK_GettingStarted_ObjectDetection.elf" to the board.
Let me know if I'm missing something here.
