RoverOne Visual Spatial Calculations using the ARToolkit


(Jaimyn Mayer) #21

Also you get a pretty abysmal framerate with computer vision tasks on a raspberry pi. Definitely recommend trying a non arm cpu like an i5 NUC or something. I blogged about a project where a friend and I used a NUC on a remote control car
here: https://jaimyn.com.au/make-autonomous-car-code-included/


(Pamela Hauff) #22

Hey Jabelone, Nice self driving remote controlled car project, I had a look. For my project the ARM processor handles the ARToolkit ok but machine learning algorithms maybe not so much. Speed doesn’t bother me because it is a research/learning project.
Though I think your right with the i5NUC they are small and low power. I will upgrade the next roverone to this computer. An i5NUC would more easily process more cameras.
There is this Gigabyte BRIX GB-BACE-3000 Barebone Kit - Celeron N3000 at $AU186.00.
It would be hard for ARM to compete with that price/cpu power.

Pam


(Stuart Longland) #23

I dunno, some of the smartphone platforms are pretty competitive with x86 from what I’ve seen. For sure, the BCM2837B0 isn’t a patch on many modern x86 chips, but then again it wouldn’t put up much of a fight against most of today’s smartphones either.

We’re talking something that’s basically level with the Snapdragon 410, which is now a 4 year old platform.

The only thing stopping these from getting up in the maker world is they’re usually glued into an impenetrable fortress of glass, aluminium and plastic, and the firmware sources/hardware design files are kept under lock-and-key.

The moment that changes (and there have been moves in that very direction), I think we’ll see a whole new class of hobbyist dev boards.


(Giles Puckett) #24

A LattePanda might be the go, if you can get hold of one.