Long-Tailed 3D Point Cloud Dataset Distillation

2026-07-29

AI Summary

This work addresses the prevalent long-tailed class distribution imbalance in 3D point cloud dataset distillation by introducing, for the first time, explicit modeling of long-tailed distributions into this task. The proposed approach features a class-adaptive synthetic budget allocation mechanism and a prior-aware supervision strategy. By aligning the global class distributions between the distilled training set and the original test set, and integrating global–local feature alignment, the method enhances the discriminability of tail classes while preserving intra-class diversity. Experiments on ShapeNet55 demonstrate that the proposed method improves classification accuracy by 7.0 percentage points over the current state-of-the-art, significantly boosting both training efficiency and model performance in long-tailed scenarios.

Abstract

Dataset distillation compresses large-scale datasets into compact synthetic sets while preserving their training utility, enabling efficient 3D point cloud training. Current point cloud dataset distillation methods only tackle geometric and representation challenges while ignoring the distributional imbalance prevalent in point cloud datasets where both training and test splits follow long-tailed class distributions. To our knowledge, we present the first study on long-tailed point cloud dataset distillation. Rather than focusing primarily on geometric and representation properties or simply constructing a class-balanced synthetic set, our framework explicitly accounts for long-tailed class distributions via two core modules. First, we design Adaptive Synthetic Budgeting to allocate class-wise synthetic budgets according to class quantity and the expected benefit of additional synthetic samples. Given the allocated budgets, we further design 3D Long-Tailed Distribution Matching to optimize synthetic point clouds through Global-Local Feature Alignment and Prior-Aware Supervision. The former preserves both global class distributions and diverse intra-class structures, while the latter provides class-dependent expert supervision to keep tail-class samples recognizable while maintaining diverse head-class patterns. Extensive experiments demonstrate the effectiveness of our method, lifting classification accuracy by 7.0 points on ShapeNet55 against state-of-the-art methods.

Problem

• long-tailed
• 3D point cloud
• dataset distillation
• class imbalance
• synthetic dataset

Innovation

• Long-tailed distribution
• Dataset distillation
• 3D point cloud
• Class imbalance
• Synthetic data