基于超点图的高斯溅写无训练层次场景理解
计算机视觉与模式识别
2025-04-18 v1
摘要
将自然语言与 3D 几何图桥接是通向灵活、语言驱动场景理解的关键步骤。虽然最近的 3D 高斯溅写 (3DGS) 进展使快速高质量场景重建成为可能，但研究也探索了将开放词汇理解整合到 3DGS 中的方法。然而，大多数现有方法需要对每个视图的 2D 语义特征图进行迭代优化，这不仅导致效率低下，也导致跨视图的语义不一致。为解决这些限制，我们引入一个无训练框架，直接从高斯原始值构建超点图。超点图将场景划分为空间紧凑且语义连贯的区域，形成一致的 3D 实体，为开放词汇理解提供结构化基础。基于图结构，我们设计一种高效的重投影策略，将 2D 语义特征提升到超点上，避免了高昂的多视图迭代训练。 resulting representation ensures strong 3D semantic coherence and naturally supports hierarchical understanding, enabling both coarse- and fine-grained open-vocabulary perception within a unified semantic field. Extensive experiments demonstrate that our method achieves state-of-the-art open-vocabulary segmentation performance, with semantic field reconstruction completed over faster. Our code will be available at https://github.com/Atrovast/THGS.
引用
@article{arxiv.2504.13153,
title = {Training-Free Hierarchical Scene Understanding for Gaussian Splatting with Superpoint Graphs},
author = {Shaohui Dai and Yansong Qu and Zheyan Li and Xinyang Li and Shengchuan Zhang and Liujuan Cao},
journal= {arXiv preprint arXiv:2504.13153},
year = {2025}
}