A study of three interfaces allowing non-expert users to teach new visual objects to a robot and their impact on learning effici