Hello, I've been using Vuforia for a few months now and am extremely impressed! Thanks for the great software. I've run across an issue that I can't find an answer to on the forums. Is there a way to not track objects that are partially occluded? Or is there a way to get back from the trackable what percent is occluded? My particular use case is this: I'm tracking a bunch of boxes with unique images. Sometimes there are mutliple boxes with the same image on the front, but when this happens they are next to each other in a nice grid.
I know vuforia won't track multiple instances of the same image target, but I thought I could get around this by taking advantage of my knowledge that when duplicates are present they will be in a specific arrangement. So I made another image target that is composed of the original image target, but in a 1x2 grid. My plan was to first check if a 'parent' image target was present (e.g. the 1x2 grid of objects) and if it was, to not render the corresponding single object image target. This has the bonus of making the 1x2 grid easier to track from a distance since it's bigger. If somebody is close enough that they don't see the whole 1x2, then the single image target shows up.
However, when I went to implement it, when looking at the single object it recognizes it as both the single image target (which is what I want) but also as the 1x2, which I assume is because Vuforia thinks the rest is just occluded. Is there a way to tell Vuforia not to track things that are largely (e.g. more than 50%) occluded, or to get the occlusion percentage for each trackable? This could be as simple as the number of points actually recognized divided by all the points, but I know this would assume a uniform distribution of points.
Thanks for any help.
Only render image targets that aren't occluded
Thanks for the update. Unfortunately Frame Markers aren't an option for me. Maybe this could be a feature in the future? From my (possible wrong) viewpoint, it wouldn't be hard to add. Even a flag that said only recognize targets where all points are visible would be helpful.