Permutation-Invariant and Orientation-Aware Dataset Distillation for 3D Point Clouds

2025-03-28

AI Summary

This work addresses the challenge of data distillation for 3D point clouds—hampered by their intrinsic permutation invariance and orientation sensitivity. We propose the first synthetic data distillation framework specifically designed for point clouds. Methodologically, we introduce a permutation-invariant feature-ranking distribution matching loss to enforce geometric structural consistency; incorporate learnable rotation parameters to adaptively align model orientations; and perform end-to-end optimization of synthetic point cloud generation. Our key contribution is the pioneering extension of data distillation to the 3D point cloud domain, overcoming both permutation and orientation constraints simultaneously. Extensive experiments on four major benchmarks—ModelNet10, ModelNet40, ShapeNet, and ScanObjectNN—demonstrate consistent and significant improvements over state-of-the-art methods, particularly in low-shot downstream classification tasks.

Technology Category

• Computer Vision: 3D Computer Vision
• Machine Learning: Learning with Manifolds
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language

Application Category

• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

We should collect large amount of data to train deep neural networks for various applications. Recently, the dataset distillation for images and texts has been attracting a lot of attention, that reduces the original dataset to a synthetic dataset while preserving essential task-relevant information. However, 3D point clouds distillation is almost unexplored due to the challenges of unordered structures of points. In this paper, we propose a novel distribution matching-based dataset distillation method for 3D point clouds that jointly optimizes the geometric structures of synthetic dataset as well as the orientations of synthetic models. To ensure the consistent feature alignment between different 3D point cloud models, we devise a permutation invariant distribution matching loss with the sorted feature vectors. We also employ learnable rotation angles to transform each syntheic model according to the optimal orientation best representing the original feature distribution. Extensive experimental results on widely used four benchmark datasets, including ModelNet10, ModelNet40, ShapeNet, and ScanObjectNN, demonstrate that the proposed method consistently outperforms the existing methods.

Problem

• Develops 3D point cloud dataset distillation method
• Ensures permutation invariance in feature alignment
• Optimizes synthetic model orientations for accuracy

Innovation

• Permutation-invariant loss for unordered point clouds
• Learnable rotation angles for optimal orientation
• Distribution matching for 3D point cloud distillation