We’re looking into getting new hardware to do AprilTags better this year, since our hardware this year was too slow to be used reliably (we got 8-10FPS with a LL2+ which because of the camera being rolling shutter, makes everything a blurry mess when you’re moving at high speeds (with photon))
I’d love to hear experiences and what setups worked for you, and recommendations for it.
First thing on my list is to try and get a global shutter camera to avoid the smearing/blurr from fast motion with low FPS.
Currently looking at maybe a Orange PI 4/5 and getting 2 global shutter cameras so we can have backward and forward coverage, would love to hear more thoughts though. since maybe just using a global shutter camera externally with a LL2+ might have better results? (We have a LL2+ & LL2)
This is the route my team plans to go. @jonahb55 and the rest of 6328 have several awesome posts about their vision system. They are the reigning ‘auto champions’ in New England partly because of the accuracy of their vision system. They used specifically the Orange Pi 5 with a camera they linked and achieved around 45-50 fps at 1600x1200.
Thats the route my team will be going this year. I’m trying to avoid spending the money on limelights as it looks like with time you can beat the performance. 6328s concept is to have multiple cameras constantly pose estimating your robot. This past year we used PhotonVision which seems a good bit less performance then leveraging OPenCVs ArUco module. But I know they were thinking of adopting that style as well.
(Edit - so sorry I can’t directly recommend this approach since we haven’t used it… YET … but I wanted to get your eyes on this one, you might find it useful)