Run machine learning demos with imx8mplus on cubox m

We have succesfully flashed the image on sd card as per the documentation :
We have logged into the cubox m, but when we check the folders like “opt” it was empty. Now we have to implement AI/ML usecases, but we do not have any samples present.

The image we are using is : sr-imx8-debian-bullseye-20210706-cli-imx8mp-sdhc-cubox-pulse.img.xz
Can you please validate that are we using the correct image ?
And for flashing the image we have used etcher tool.
Following are some screenshots of image flashed in sd card:


Our Debian image is a base image that does not support any of the AI acceleration features provided in NXP’s BSP. Please use one of the prebuilt Yocto images for developing and testing the AI acceleration features on the iMX8MP.

Can you please provide me the source from where the yocto based image can be downloaded? Also can you tell the specific image if multiple images can be found?


The Yocto source can be found here with instructions, GitHub - SolidRun/meta-solidrun-arm-imx8 I recommend using the newest released kirkstone branch as that is an LTS yocto release and has the latest NXP BSP for ML integrated into it.

We do not have prebuilt kirkstone images yet, but you can test with the previous release here, SolidRun Images

From the link you shared, we used the following image: imx-image-full-imx8mpsolidrun.wic.bz2
After flashing this image on sd card and then putting the sd card back in device, we got directory structure as follows:

After checking all the folders, we could not find any ML examples. Can you please help where to find them or provide us with another yocto image that is supported on the device?


You need to refer to NXP’s BSP documentation on their machine learning examples.

Alternatively you can refer to the demo that I put together over on the DevCuts youtube channel.