- Edited
I'm a mentor for a First Robotics Competition (FRC) team. The fields for FRC are thoroughly April-Tagged, and most teams either use Limelight cameras or COTS cameras with PhotonVision for localization by AprilTag. The former always has supply issues, and our team discovered that you need a large amount of time, plus access to a full field, to properly calibrate cameras for the latter. We're a small team with limited space and access to practice fields, so I am excited about the possibility of using object + depth detection to get the "down to the inch" precision that successful teams have.
That said, being able to localize (even with low precision) from AprilTags is critical, and given power & weight I don't want to have two different vision systems running simultaneously. I saw that current models aren't able to process AprilTags efficiently, so I'm curious about a couple things:
- Is RVC4 turning out to be a better fit for AprilTags? Will it have better support?
- If so, are RVC4 devices in a similar price range to the OAK-D Pro going to be available in the coming months?
- If not, is there a way to get the depth perception & object detection from an OAK-D device, but simultaneously pass the original camera feed to another coprocessor to handle AprilTag localization?