Any info on using the BeagleBone AI's Embedded Vision Engines?

Question

The Beaglebone AI looks amazing, but I can't seem to find any info on making use of the 4 Embedded Vision Engines in the Sitara AM5729. Does anyone know of any good quick starts or examples?

I've found a few things linked from this page, but that's all. SITARA-MACHINE-LEARNING Sitara Machine Learning | TI.com

It took quite some effort to get up to date information on the PRUs. I suspect this may be similar!

shabaz · Accepted Answer

Hi Fred,

There's a few links here Another BeagleBone - BeagleBone AI! when there were initial reports of the BBB-AI, but like you, I really hope that it's all much more mature than the PRU documentation was on day-one! The EVE functionality looks extremely attractive for accelerating a ton of image-related stuff, I hope to find time to learn more about it too.

Also, if you're not subscribed to the BeagleBoard mailer, it will be worth doing so, because useful information appears there too (although it's noisy, and often awkward to follow, since it is old-school text mailer format).

Fred27 · Answer

This seems useful too: https://www.hackster.io/175809/tidl-on-beaglebone-ai-1ee263

It seems that some of the TIDL (TI Deep Learning) part of the Sitara Processor SDK has been ported into the Beaglebone AI Debian image. It looks like TIDL is the sensible way to use the EVE and DSP components rather than trying to do things directly.

Any info on using the BeagleBone AI's Embedded Vision Engines?

Top Replies