We’re looking for feedback from teams whom have previously used/currently using the Arducam OV9281 Usb camera with an OrangePi 5 coprocessor to run PhotonVision Apriltag detection. Are there any suggestions/key areas to look out for you guys can give us? Much Thanks!!
Our team tested that exact setup.
We got around 60 fps consistently with very good results. Some of the exposure settings acted weird, but I believe this was solved at some point by pv.
Imo, this is the best vision setup for detecting tags. Either that or an opi 5 with OV2311.
1 Like
That’s really good to hear, did you use a separate camera for the driver or just streamed the output from OV9281?
We have another camera for object detection so naturally we chose that one
I assume drivers would rather have color in the camera stream
Totally forgot the arducam was mono… What was the model of the other camera you used? Also did you guys connected it through photon vision with the arducam or directly to roborio?
In general if a coproc is possible I’d never use roborio for camera things directly
2 Likes
We didn’t fully have this setup in comp as we got our cameras around the time of the preseason
Rn, we have created a photon vision fork to run ml on the rockchip so both cameras will be connected to the same orange pi. I’ll try to remember later today sending the results of having multiple camera’s connected and running ml
1 Like
Just tested
I got easily 50~ fps with the OV9281 while maintaining fast updates and long distance detections. (With the rknn ml running)
This is just touch and go numbers, can definitely be optimized
1 Like
That’s great to hear. The discord invite link for the beta testing seems to be down on the thread you shared, could you perhaps update it? Thanks!
Yeah sorry😅
I put the expiring link in the start and now I can’t edit it anymore
The updated link is here