By njessenberg
on Thu, 08/29/2019 - 13:49
Hello,
Currently we are using MT360 for recognition of objects. However, in some cases, objects can be similar to the point where Vuforia recognizes objects when it should not.
My question is whether I can control the threshold within Vuforia, or at least query the precision of the current recognition?
I realize higher thresholds might complicate recognition, but we would like to experiment with different values to see if this can help us. Or at the very least we can be more cautious when Vuforia is only 0.x certain of its current detection, if there is some way to query those values.
Thanks :)
Hi,
Hi,
False positive detections are a known issue, in particular keyboards or other extremely cluttered areas are good target traps. This also happens more often when the Model Target has simple geometry, or when no material information is given in the 3D model.
Thank you for the information
Thank you for the information.
Yes, unfortunately, our objects do not have much in terms of special textures, just a single dark color. The mostly have sharp edges luckily and some other features, but that is mostly it.