3DMatch: Learning Local Geometric Descriptors from RGB-D Reconstructions

TL;DR

3DMatch introduces a data-driven local geometric descriptor learned from RGB-D reconstructions, significantly improving matching accuracy on noisy, incomplete 3D data and generalizing across tasks and scales.

Contribution

The paper presents a novel self-supervised learning approach for local 3D feature descriptors that outperforms existing methods and generalizes well across various applications.

Findings

Outperforms state-of-the-art descriptors in 3D matching tasks

Generalizes to different tasks like object alignment and surface correspondence

Effective on noisy, incomplete real-world 3D scan data

Abstract

Matching local geometric features on real-world depth images is a challenging task due to the noisy, low-resolution, and incomplete nature of 3D scan data. These difficulties limit the performance of current state-of-art methods, which are typically based on histograms over geometric properties. In this paper, we present 3DMatch, a data-driven model that learns a local volumetric patch descriptor for establishing correspondences between partial 3D data. To amass training data for our model, we propose a self-supervised feature learning method that leverages the millions of correspondence labels found in existing RGB-D reconstructions. Experiments show that our descriptor is not only able to match local geometry in new scenes for reconstruction, but also generalize to different tasks and spatial scales (e.g. instance-level object model alignment for the Amazon Picking Challenge, and mesh surface correspondence). Results show that 3DMatch consistently outperforms other state-of-the-art approaches by a significant margin. Code, data, benchmarks, and pre-trained models are available online at http://3dmatch.cs.princeton.edu