SuperFlex: Deformable Superquadrics for Point Cloud Decomposition

2026-07-01

AI Summary

Existing superquadric-based methods for point cloud reconstruction suffer from low geometric accuracy, support only rigid primitives, and exhibit poor robustness to partial observations. To address these limitations, this work proposes SuperFlex, a novel framework that, for the first time, incorporates deformable mechanisms—such as bending and tapering—into superquadric representations. A new reconstruction loss function is introduced to enhance geometric fidelity, and high-quality decomposition results are leveraged to supervise deep network training, significantly improving robustness to partial point clouds in real-world scenes. The proposed approach achieves substantially higher reconstruction accuracy than current optimization- and learning-based baselines while maintaining highly compact primitive representations.

Abstract

Superquadrics have proven to provide a compact, geometrically meaningful representation for 3D objects. However, existing methods suffer from limited reconstruction accuracy, are restricted to rigid primitives, and lack robustness to partial point clouds. In this work, we present SuperFlex, an enhanced framework that expands the expressive power and applicability of superquadric decompositions. First, we introduce a novel loss formulation which significantly improves reconstruction accuracy. Second, we include bending and tapering deformations, enabling high-fidelity representation of curved and asymmetric geometries. Finally, we leverage these high-quality decompositions as supervision to train a model that is robust to partial real-world point clouds. Experiments demonstrate substantial improvements in reconstruction accuracy over both optimization- and learning-based baselines while maintaining a highly compact primitive representation.

Problem

• superquadrics
• point cloud decomposition
• reconstruction accuracy
• partial point clouds
• geometric representation

Innovation

• superquadrics
• deformable primitives
• point cloud decomposition
• bending and tapering
• robust reconstruction