The lack of 3D understanding in computer vision systems is due to the failure of techniques that use multi view geometry when there is only a single view of a scene.