Does ORB SLAM2 perform worse on weaker compute?!
Hi Everyone,
I'm working on trying to get ORB SLAM 2 to work well on a Raspberry Pi 4 and have a few ideas on how to do that. Iv'e cloned the official Git repo built it successfully and am doing test runs on on the Kitti data set.
But I don't get the good result that I'm seeing in Raul's videos running on the same data set.
Does anyone know if there's anything that make the code perform with worse quality on weaker compute? I would have thought that it should still do all the computations it needs and just run for longer, when pulling image stream from disk. (so it's not forced to a particular frame rate)
Cheers Fred
what exactly are you refering to ?
When I look at Raul's video its performing really well, here https://www.youtube.com/watch?v=51NQv...
When I'm running the built official project, from here https://github.com/raulmur/ORB_SLAM2
on the exact same data set as in that video I'm getting nowhere near the same quality of tracking and map build. When I run it looses the track many times and doesn't consistently add points to the map.
@RedMarsBlueMoon is your previous question solved?
(REGARDING above mentioned 'Previous question') I think I ended up doing that other thing slightly differently. It was a while ago now. We can set it to resolved?
i don't think, we can help you debug a whole 3rdparty repo.
try to make an issue there
Sure I'll continue with this in a different way. I just thought that since this algo is now part of OpenCV that someone might have a good general understanding of how it works. But no worries I'll try somewhere else.
no, it's not