You could project the 8 vertices of the bounding box and take the min/max of x/y for a crude 2D bounding rectangle. If you need more precision and pixel perfect collision detection, then I don’t see an easy way to do that… I think the best way would still be to have your objects in 3D with correct z positionning for the regular collision system to work. If that is a problem for rendering order that the z-buffer can’t handle, maybe split some objects to dispatch each parts in different rendering groups?