Learning object affordances by leveraging the combination of human-guidance and self-exploration