G2P: Gaussian-to-Point Attribute Alignment for Boundary-Aware 3D Semantic Segmentation

2026-01-07

arXiv.org

AI Summary

This work addresses the challenge in point cloud semantic segmentation where geometrically similar but visually distinct objects—differing in appearance such as color or material—are difficult to discriminate. To resolve this, the authors propose transferring appearance-aware attributes from 3D Gaussian splatting to the original point cloud through a Gaussian-to-point alignment mechanism. By leveraging Gaussian opacity to alleviate geometric ambiguity and utilizing its scale parameters for precise boundary localization, the method achieves enhanced segmentation accuracy without relying on 2D images or language supervision. Evaluated on standard benchmarks, the approach outperforms existing methods, particularly excelling on geometrically complex or challenging categories by significantly improving both segmentation precision and appearance consistency.

Technology Category

• Computer Vision: Segmentation
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Machine Learning: Hardware-aware ML

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Security and Privacy: Data transparency and provenance
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Semantic segmentation on point clouds is critical for 3D scene understanding. However, sparse and irregular point distributions provide limited appearance evidence, making geometry-only features insufficient to distinguish objects with similar shapes but distinct appearances (e.g., color, texture, material). We propose Gaussian-to-Point (G2P), which transfers appearance-aware attributes from 3D Gaussian Splatting to point clouds for more discriminative and appearance-consistent segmentation. Our G2P address the misalignment between optimized Gaussians and original point geometry by establishing point-wise correspondences. By leveraging Gaussian opacity attributes, we resolve the geometric ambiguity that limits existing models. Additionally, Gaussian scale attributes enable precise boundary localization in complex 3D scenes. Extensive experiments demonstrate that our approach achieves superior performance on standard benchmarks and shows significant improvements on geometrically challenging classes, all without any 2D or language supervision.

Problem

• 3D semantic segmentation
• point cloud
• appearance ambiguity
• geometric ambiguity
• boundary-aware segmentation

Innovation

• Gaussian-to-Point
• 3D Gaussian Splatting
• point cloud segmentation
• boundary-aware
• appearance alignment