Unified Primitive Proxies for Structured Shape Completion

2026-01-02

arXiv.org

AI Summary

This work addresses the challenge of recovering structured geometric shapes from incomplete 3D data by proposing an end-to-end shape completion method that departs from conventional cascaded pipelines. The approach introduces learnable primitive proxies and a dedicated attention-based decoding pathway to jointly predict geometry, semantics, and inlier membership within a unified framework. Through an online target update strategy, primitive proxies and point clouds are co-optimized to yield assembly-ready, structured representations. Evaluated across multiple synthetic and real-world benchmarks and integrated with four distinct assembly solvers, the method achieves up to a 50% reduction in Chamfer distance and improves normal consistency by as much as 7%.

Technology Category

Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal ReasoningComputer Vision: 3D Computer VisionHumans and AI: Human-Aware Planning and Behavior Prediction

Application Category

Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMsGraph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphsEconomics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI

Abstract

Structured shape completion recovers missing geometry as primitives rather than as unstructured points, which enables primitive-based surface reconstruction. Instead of following the prevailing cascade, we rethink how primitives and points should interact, and find it more effective to decode primitives in a dedicated pathway that attends to shared shape features. Following this principle, we present UniCo, which in a single feed-forward pass predicts a set of primitives with complete geometry, semantics, and inlier membership. To drive this unified representation, we introduce primitive proxies, learnable queries that are contextualized to produce assembly-ready outputs. To ensure consistent optimization, our training strategy couples primitives and points with online target updates. Across synthetic and real-world benchmarks with four independent assembly solvers, UniCo consistently outperforms recent baselines, lowering Chamfer distance by up to 50% and improving normal consistency by up to 7%. These results establish an attractive recipe for structured 3D understanding from incomplete data. Project page: https://unico-completion.github.io.

Problem

• structured shape completion
• 3D reconstruction
• geometric primitives
• incomplete data
• shape understanding

Innovation

• primitive proxies
• structured shape completion
• unified representation
• assembly-ready output
• online target updates