Fully Self-Supervised Class Awareness in Dense Object Descriptors

TL;DR

This paper presents a self-supervised approach for learning dense, class-aware object descriptors that improve pixel-level correspondence accuracy in multi-object scenes, with applications in robotic grasping.

Contribution

It introduces a novel self-supervised method for dense object descriptors that incorporate class-awareness, outperforming previous methods in robustness and accuracy.

Findings

Outperforms previous techniques in dense correspondence accuracy

Provides robust pixel-to-pixel matching in cluttered scenes

Enables robotic grasping based on dense correspondences

Abstract

We address the problem of inferring self-supervised dense semantic correspondences between objects in multi-object scenes. The method introduces learning of class-aware dense object descriptors by providing either unsupervised discrete labels or confidence in object similarities. We quantitatively and qualitatively show that the introduced method outperforms previous techniques with more robust pixel-to-pixel matches. An example robotic application is also shown~- grasping of objects in clutter based on corresponding points.