Deep Weighted Consensus: Dense correspondence confidence maps for 3D shape registration

TL;DR

This paper introduces a robust, differentiable method for rigid point cloud alignment that effectively handles large rotations and noise, outperforming existing models in accuracy and robustness.

Contribution

The authors propose a novel dense confidence map-based approach for 3D shape registration that is robust to noise and large rotations, unlike prior constrained models.

Findings

Outperforms recent methods like DCP, PointNetLK, RPM-Net, PRnet

Effective under large rotations and high noise levels

Converges with high accuracy in full SO(3) spectrum

Abstract

We present a new paradigm for rigid alignment between point clouds based on learnable weighted consensus which is robust to noise as well as the full spectrum of the rotation group. Current models, learnable or axiomatic, work well for constrained orientations and limited noise levels, usually by an end-to-end learner or an iterative scheme. However, real-world tasks require us to deal with large rotations as well as outliers and all known models fail to deliver. Here we present a different direction. We claim that we can align point clouds out of sampled matched points according to confidence level derived from a dense, soft alignment map. The pipeline is differentiable, and converges under large rotations in the full spectrum of SO(3), even with high noise levels. We compared the network to recently presented methods such as DCP, PointNetLK, RPM-Net, PRnet, and axiomatic methods such as ICP and Go-ICP. We report here a fundamental boost in performance.