Object recognition using visuo-affordance maps