Hello!
BIOBUZZ’s new AprilTag cluster system sounds like it will improve accuracy a lot, but I’m not sure how each AprilTagProcessor.PoseSolver will interact with it.
If you’ve tested clusters, how has your PoseSolver been working for you?
Also, does IPPE_SQUARE work well with them? Or do they count as rectangles?
What behavior are you seeing? Can you provide a few examples?
I haven’t been able to test the new cluster system yet, so I wanted to see how it’s going for other people. I’m trying to figure out how the cluster tags affect pose solving.
It shouldn’t affect pose solving at all. If it does, let us know and we will fix it quickly.
We released the production art for the AprilTags yesterday, you should be able to download them and point your webcam to the screen to do some basic testing. I think we should be able to release printable versions within the next week at this same location:
Do you know if clusters are treated as 4 individual AprilTags, just with their data combined, or as 1 big AprilTag? If it’s the latter, will IPPE_SQUARE still work?
This AprilTag Clusters resource should help you understand the AprilTag Clusters better, it’s a tech tip that describes how AprilTag Clusters are managed by the SDK.
The AprilTag Cluster algorithm doesn’t just use four corners from a single AprilTag, it uses all of the corners of all detected AprilTags (that are members of the cluster) to provide pose information, making the cluster much more accurate than a single target. The only difference between the Single AprilTag and the AprilTag Clusters is the metadata provided in the AprilTagSingleDetection and AprilTagClusterDetection classes, and I highlighted those differing methods at the bottom of the AprilTag Clusters tech tip.
Oh, okay, thank you for explaining!