Mapping Visual Behavior to Robotic Assembly Tasks