Improving vision-based robotic manipulation with affordance understanding