0
$\begingroup$

I am building a robot which can move in multiple terrains using legs. I have decided to use nvidia jetson xavier nx. Is this board capable of running the following

  1. Yolov5 for real time object detection.( I already tried it in raspberry pi but it was no way near to working as the frames were rendered really slow)
  2. Real time 3D-reconstructio from stereo vision. (I am yet to find a solution for this, but I want to decide my SBC first so that I wont run into complications like it was for yolov5. I am trying to do something similiar to this (video link) or similiar to what tesla did for 4D auto navigation.
  3. For Future scalability, I want it to be capable of processing some self learning approaches and has some AI.

As from my research I found xavier nx was one of the most powerful boards, will it be capable of processing these things in and around 30 - 60 fps, or should I go for jetson AGX Xavier.

I dont want to do any remote processing as latency is a key constraint. So I want to build a standalone robot. Also looking for battery solutions wrt the board. Currently for powering up the rpi I am using a powerbank which should be enough to deliver 15W. But as the boards performance increase there will be a demand for power. The budget is not a concern.

Thanks in advance.

$\endgroup$
1

1 Answer 1

0
$\begingroup$

The xavier nx, from personal experience, has more than enough juice for yolov5 and I would also assume it's more than capable of handling your other requirements. Pair it with a high-capacity 3S or 4S lipo (with appropriate circuitry), you should be able to power it for way longer than a standard power bank. If budget is not a concern for you, have you considered the orin series? I believe the orin nx is the same form factor as the xavier nx, but ~3.5-5x better.

$\endgroup$

Not the answer you're looking for? Browse other questions tagged or ask your own question.