SurfR: Surface Reconstruction with Multi-scale Attention

2025-06-10

AI Summary

Existing point cloud surface reconstruction methods face a fundamental accuracy-efficiency trade-off: small, object-specific models achieve high accuracy but suffer from poor generalization and require per-instance training; conversely, general-purpose large models generalize well yet lack fine geometric detail and incur high inference latency. This paper proposes a universal implicit representation framework for unstructured point clouds, introducing a novel *lazy query* mechanism that skips redundant spatial sampling. Coupled with parallel multi-scale voxel grid representations and cross-scale attention-based feature fusion, the framework enables efficient hierarchical feature aggregation. Our method achieves near-state-of-the-art reconstruction accuracy while significantly outperforming all baselines—including those using optimal resolution settings—in inference speed. Moreover, it demonstrates substantially improved robustness to input noise and enhanced adaptability to varying input point densities.

Technology Category

• Computer Vision: Representation Learning for Vision
• Machine Learning: Feature Construction/Reformulation
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies

Abstract

We propose a fast and accurate surface reconstruction algorithm for unorganized point clouds using an implicit representation. Recent learning methods are either single-object representations with small neural models that allow for high surface details but require per-object training or generalized representations that require larger models and generalize to newer shapes but lack details, and inference is slow. We propose a new implicit representation for general 3D shapes that is faster than all the baselines at their optimum resolution, with only a marginal loss in performance compared to the state-of-the-art. We achieve the best accuracy-speed trade-off using three key contributions. Many implicit methods extract features from the point cloud to classify whether a query point is inside or outside the object. First, to speed up the reconstruction, we show that this feature extraction does not need to use the query point at an early stage (lazy query). Second, we use a parallel multi-scale grid representation to develop robust features for different noise levels and input resolutions. Finally, we show that attention across scales can provide improved reconstruction results.

Problem

• Fast surface reconstruction for unorganized point clouds
• Balancing detail accuracy and generalization in 3D shapes
• Improving speed and performance with multi-scale attention

Innovation

• Lazy query for faster feature extraction
• Parallel multi-scale grid representation
• Attention across scales improves reconstruction