Error while converting the yolov8n.onnx model

I use Linux kernel version 6.0.0 with NPU patches

root@khadas-vim3:~# uname -a
Linux khadas-vim3 6.0.0-yocto-standard #1 SMP Sun Oct 2 21:09:07 UTC 2022 aarch64 GNU/Linux

Hello @aananthcn ,

I run your model on VIM3 with kernel vim3-ubuntu-22.04-gnome-linux-6.0-fenix-1.3-221118.img.xz. But nothing wrong.

Hello @Louis-Cheng-Liu, than you very much for your support, I am able to run the app and the model on VIM3 with Yocto baseline. And I can confirm that it is using the NPU (the load average is less than 1 core) to detect face/objects.

Following are the corrections:

  1. The version of the libraries (listed below) that I was using in my Yocto baseline were old. I now upgraded them to the one that is available in the image link you shared above.
    libs => libArchModelSw.so, libCLC.so, libGAL.so, libNNArchPerf.so, libOpenVX.so, libOpenVXU.so, libovxlib.so, libVSC.so

  2. The way I was building my application using Yocto recipe was wrong, it did not include the environment variable settings that are set in build_vx.sh build script.

1 Like