Curvature Diversity-Driven Deformation and Domain Alignment for Point Cloud

TL;DR

This paper introduces a novel unsupervised domain adaptation method for point clouds, leveraging curvature-based deformation and nuclear-norm Wasserstein discrepancy to improve domain alignment and achieve state-of-the-art results.

Contribution

The paper proposes a new domain adaptation framework combining curvature-driven deformation reconstruction and nuclear-norm Wasserstein discrepancy, with theoretical validation and extensive experiments.

Findings

Achieves state-of-the-art performance on point cloud domain adaptation tasks.

Effectively aligns source and target domains using curvature-based deformations.

Demonstrates the generality of D-NWD for various deformations.

Abstract

Unsupervised Domain Adaptation (UDA) is crucial for reducing the need for extensive manual data annotation when training deep networks on point cloud data. A significant challenge of UDA lies in effectively bridging the domain gap. To tackle this challenge, we propose \textbf{C}urvature \textbf{D}iversity-Driven \textbf{N}uclear-Norm Wasserstein \textbf{D}omain Alignment (CDND). Our approach first introduces a \textit{\textbf{Curv}ature Diversity-driven Deformation \textbf{Rec}onstruction (CurvRec)} task, which effectively mitigates the gap between the source and target domains by enabling the model to extract salient features from semantically rich regions of a given point cloud. We then propose \textit{\textbf{D}eformation-based \textbf{N}uclear-norm \textbf{W}asserstein \textbf{D}iscrepancy (D-NWD)}, which applies the Nuclear-norm Wasserstein Discrepancy to both \textit{deformed and original} data samples to align the source and target domains. Furthermore, we contribute a theoretical justification for the effectiveness of D-NWD in distribution alignment and demonstrate that it is \textit{generic} enough to be applied to \textbf{any} deformations. To validate our method, we conduct extensive experiments on two public domain adaptation datasets for point cloud classification and segmentation tasks. Empirical experiment results show that our CDND achieves state-of-the-art performance by a noticeable margin over existing approaches.