So does this split a partition if there are a lot of objects in a partition? Or is a space only split like in the pattern shown above?
It seems like there is a set location for each of the splits. Would it be possible to split on a slightly different ratio eat layer?
I wonder if there is a way to automatically detect what is the best spacial sturcture given a triangle distribution?
I remember it was mentioned this isn't used much in practice, but it seems here like it could be quite useful. Is there something that incurs a high cost about this method?
Do we always split the quadrants, or only in areas where the objects are highly concentrated?
How does the code to determine this split different from that of k-d trees?
Is the disadvantage compared to K-D Tree caused by inflexible division point decision?