InvariantOODG: Learning Invariant Features of Point Clouds for Out-of-Distribution Generalization
Zhimin Zhang, Xiang Gao, Wei Hu
TL;DR
Addressing the challenge of 3D point-cloud generalization across distribution shifts, the work introduces InvariantOODG, a two-branch framework that processes the original and augmented point clouds with a shared feature extractor to learn invariant representations. It keyly integrates a dynamic anchor-point module for robust local region matching and enforces local and global invariance through Chamfer-based alignment and pooling-based losses, formalized with $\ ext{L}_{CD}$, $\text{L}_{local}$, and $\text{L}_{global}$ terms. The approach is validated on sim-to-real and PointDA benchmarks, achieving state-of-the-art or competitive results and demonstrating clear gains in OOD robustness. Overall, the method enhances the reliability of point-cloud understanding under distributional shifts, enabling more practical deployment in real-world sensing tasks, with performance supported by end-to-end optimization of the combined loss across multiple scales and regions, illustrated by comparisons to baselines such as MetaSets and PDG.
Abstract
The convenience of 3D sensors has led to an increase in the use of 3D point clouds in various applications. However, the differences in acquisition devices or scenarios lead to divergence in the data distribution of point clouds, which requires good generalization of point cloud representation learning methods. While most previous methods rely on domain adaptation, which involves fine-tuning pre-trained models on target domain data, this may not always be feasible in real-world scenarios where target domain data may be unavailable. To address this issue, we propose InvariantOODG, which learns invariability between point clouds with different distributions using a two-branch network to extract local-to-global features from original and augmented point clouds. Specifically, to enhance local feature learning of point clouds, we define a set of learnable anchor points that locate the most useful local regions and two types of transformations to augment the input point clouds. The experimental results demonstrate the effectiveness of the proposed model on 3D domain generalization benchmarks.