Learning to See before Learning to Act: Visual Pre-training for Manipulation

TL;DR

Pre-training visual models significantly enhances robotic manipulation capabilities, enabling zero-shot object pickup and improving sample efficiency through transfer learning of affordance-related features.

Contribution

The paper demonstrates that transferring features from vision models to manipulation tasks enables zero-shot learning and improves sample efficiency in robotic object manipulation.

Findings

Zero-shot object pickup with no robotic experience.

80% success rate after minimal robotic training.

Transfer of visual features enhances manipulation performance.

Abstract

Does having visual priors (e.g. the ability to detect objects) facilitate learning to perform vision-based manipulation (e.g. picking up objects)? We study this problem under the framework of transfer learning, where the model is first trained on a passive vision task, and adapted to perform an active manipulation task. We find that pre-training on vision tasks significantly improves generalization and sample efficiency for learning to manipulate objects. However, realizing these gains requires careful selection of which parts of the model to transfer. Our key insight is that outputs of standard vision models highly correlate with affordance maps commonly used in manipulation. Therefore, we explore directly transferring model parameters from vision networks to affordance prediction networks, and show that this can result in successful zero-shot adaptation, where a robot can pick up certain objects with zero robotic experience. With just a small amount of robotic experience, we can further fine-tune the affordance model to achieve better results. With just 10 minutes of suction experience or 1 hour of grasping experience, our method achieves ~80% success rate at picking up novel objects.