Hi-TOPS: Hierarchical Topology-aware Scoring Prior for 3D Part Decomposition

2026-08-01

AI Summary

Existing unsupervised 3D part decomposition methods struggle to simultaneously achieve precise articulation boundary alignment and robustness to noise at the structural scale. This work proposes a hierarchical, topology-aware scoring prior that introduces, for the first time, a multi-resolution Flow-Freeze field to integrate intrinsic geometric cues. By combining TSDF-guided superquadric fitting with connectivity-aware mesh assignment, our approach yields structurally plausible, boundary-aligned, and editable decompositions. Notably, the method operates without semantic supervision or 2D priors, and consistently outperforms current unsupervised techniques across multiple benchmarks, demonstrating superior stability, accuracy, and editability.

Abstract

Accurate 3D part decomposition requires separating shapes into structurally meaningful components with precise boundaries while preserving articulation seams and thin attachments. Existing approaches often suffer from a structural-scale mismatch: geometric evidence for separation is most reliable at the meso scale, yet many pipelines operate either too globally to respect joints or too locally to remain robust to noise. We propose Hi-TOPS, a Hierarchical Topology-aware Scoring Prior that aggregates complementary intrinsic cues into a multi-resolution Flow-Freeze field. Flow regions provide expandable support for primitive coverage, while Freeze regions restrict growth near articulations and thin structures. A TSDF-guided body-surface superquadric fitter then captures dominant cores and residual surface structures, followed by SQ-to-mesh assignment for connected, boundary-aligned parts. Across diverse benchmarks, Hi-TOPS delivers stable, editable decompositions without semantic supervision or 2D foundation priors.

Problem

• 3D part decomposition
• structural-scale mismatch
• articulation seams
• thin attachments
• boundary alignment

Innovation

• Hierarchical Topology-aware Scoring
• Flow-Freeze Field
• TSDF-guided Superquadric Fitting
• 3D Part Decomposition
• Boundary-aligned Mesh Assignment