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

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Song, Hojun, Song, Chae-yeong, Hong, Jeong-hun, Moon, Chaewon, Kim, Dong-hwi, Kim, Gahyeon, Kim, Soo Ye, Liao, Yiyi, Lee, Jaehyup, Park, Sang-hyo
Format: Preprint
Publié: 2026
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866915783598145536
author Song, Hojun
Song, Chae-yeong
Hong, Jeong-hun
Moon, Chaewon
Kim, Dong-hwi
Kim, Gahyeon
Kim, Soo Ye
Liao, Yiyi
Lee, Jaehyup
Park, Sang-hyo
author_facet Song, Hojun
Song, Chae-yeong
Hong, Jeong-hun
Moon, Chaewon
Kim, Dong-hwi
Kim, Gahyeon
Kim, Soo Ye
Liao, Yiyi
Lee, Jaehyup
Park, Sang-hyo
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2601_03510
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle G2P: Gaussian-to-Point Attribute Alignment for Boundary-Aware 3D Semantic Segmentation
Song, Hojun
Song, Chae-yeong
Hong, Jeong-hun
Moon, Chaewon
Kim, Dong-hwi
Kim, Gahyeon
Kim, Soo Ye
Liao, Yiyi
Lee, Jaehyup
Park, Sang-hyo
Computer Vision and Pattern Recognition
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.
title G2P: Gaussian-to-Point Attribute Alignment for Boundary-Aware 3D Semantic Segmentation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2601.03510