Symmetria: A Synthetic Dataset for Learning in Point Clouds

2025-10-27

AI Summary

Point cloud learning is hindered by the scarcity of large-scale labeled datasets. To address this, we introduce Symmetria—the first scalable point cloud dataset systematically constructed from mathematical symmetry principles. Our method explicitly models symmetry groups to synthesize highly diverse point clouds with ground-truth symmetry annotations, enabling arbitrary-scale generation. We further define and benchmark “symmetry detection” as a novel task, establishing the first standardized evaluation protocol. Empirically, Symmetria significantly improves performance across self-supervised pretraining, few-shot learning, and real-world fine-tuning, demonstrating strong generalization on downstream tasks including classification and segmentation. Pretrained models leveraging Symmetria achieve an average accuracy gain of +4.2% on real-world benchmarks such as ScanObjectNN. By providing a reproducible, extensible, and theoretically grounded data infrastructure, Symmetria advances foundational research in point cloud representation learning.

Technology Category

Computer Vision: Visual Reasoning & Symbolic RepresentationsMachine Learning: Neuro-Symbolic LearningNatural Language Processing: Code Generation / Program Synthesis from Natural Language

Application Category

Security and Privacy: Data transparency and provenanceSemantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologiesWeb Mining and Content Analysis: Large pretrained models with web data

Abstract

Unlike image or text domains that benefit from an abundance of large-scale datasets, point cloud learning techniques frequently encounter limitations due to the scarcity of extensive datasets. To overcome this limitation, we present Symmetria, a formula-driven dataset that can be generated at any arbitrary scale. By construction, it ensures the absolute availability of precise ground truth, promotes data-efficient experimentation by requiring fewer samples, enables broad generalization across diverse geometric settings, and offers easy extensibility to new tasks and modalities. Using the concept of symmetry, we create shapes with known structure and high variability, enabling neural networks to learn point cloud features effectively. Our results demonstrate that this dataset is highly effective for point cloud self-supervised pre-training, yielding models with strong performance in downstream tasks such as classification and segmentation, which also show good few-shot learning capabilities. Additionally, our dataset can support fine-tuning models to classify real-world objects, highlighting our approach's practical utility and application. We also introduce a challenging task for symmetry detection and provide a benchmark for baseline comparisons. A significant advantage of our approach is the public availability of the dataset, the accompanying code, and the ability to generate very large collections, promoting further research and innovation in point cloud learning.

Problem

• Addresses point cloud dataset scarcity through synthetic generation
• Enables data-efficient learning with precise ground truth availability
• Supports generalization across geometric tasks and real-world applications

Innovation

• Synthetic dataset generation via symmetry formulas
• Enables self-supervised pre-training for point clouds
• Supports extensible tasks with precise ground truth