Embodied Uncertainty-Aware Object Segmentation

TL;DR

This paper presents UncOS, a novel uncertainty-aware object segmentation method that improves robot perception by generating multiple segmentation hypotheses with confidence estimates, enabling better scene understanding and interaction.

Contribution

The paper introduces a new approach for generating a distribution of segmentation hypotheses with confidence, enhancing embodied robot perception and interaction capabilities.

Findings

Achieves state-of-the-art performance on unseen object segmentation tasks.

Enables belief-driven robot actions to reduce scene ambiguity.

Demonstrates effectiveness in real-robot experiments.

Abstract

We introduce uncertainty-aware object instance segmentation (UncOS) and demonstrate its usefulness for embodied interactive segmentation. To deal with uncertainty in robot perception, we propose a method for generating a hypothesis distribution of object segmentation. We obtain a set of region-factored segmentation hypotheses together with confidence estimates by making multiple queries of large pre-trained models. This process can produce segmentation results that achieve state-of-the-art performance on unseen object segmentation problems. The output can also serve as input to a belief-driven process for selecting robot actions to perturb the scene to reduce ambiguity. We demonstrate the effectiveness of this method in real-robot experiments. Website: https://sites.google.com/view/embodied-uncertain-seg