Fixating on one object at a time helps robustly recover 3D information relevant to interacting with that object.