Does anyone know what is the current state of 3D object recognition using Shapenet for example? I imagine there must be progress on this in the AR industry.