LangRef3DGS：基于 3D 高斯泼溅、由自然语言引导的部分观测三维指代分割
LangRef3DGS: Natural Language-Guided 3D Referential Segmentation from Partial Observations via 3D Gaussian Splatting
把 3DGS 当作可微的语义嵌入场来承载开放词表文本查询，并用梯度低秩机制压缩特征冗余以提升类别可分性——这套「嵌入场 + 低秩」组合可直接迁移到任何需要从稀疏 RGB-D 观测做语言查询 3D 分割或新类发现的场景。
原文依据 a gradient low-rank mechanism that enhances class separability by reducing feature redundancy
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论语言引导的三维分割对于连接三维感知与语义理解至关重要，然而在真实 RGB-D 数据中常见的稀疏与遮挡视角下，它仍然十分脆弱。为克服这一问题，我们提出一个实时框架，利用 3D 高斯泼溅（3DGS）从部分观测中构建语义连续且可微的嵌入场。我们的方法集成了两个关键组件：用于自适应发现新物体类别的狄利克雷过程（DP），以及通过降低特征冗余来增强类别可分性的梯度低秩机制。二者的结合使得在文本提示直接引导下实现鲁棒的开放词表分割成为可能。在具有挑战性的基准上的大量实验表明，我们的方法取得了强劲的性能，展现出更高的准确率、对部分输入的鲁棒性以及强大的新类发现能力。我们的代码与模型可在 https://github.com/Tap12345/LangGS 获取。
英文原文摘要 986 字符
Language-guided 3D segmentation is crucial for linking 3D perception with semantic understanding, yet it remains vulnerable to the sparse and occluded views common in real-world RGB-D data. To overcome this, we present a real-time framework that leverages 3D Gaussian Splatting (3DGS) to build a semantically continuous and differentiable embedding field from partial observations. Our approach integrates two key components: a Dirichlet Process (DP) for the adaptive discovery of novel object categories, and a gradient low-rank mechanism that enhances class separability by reducing feature redundancy. This combination enables robust open-vocabulary segmentation guided directly by text prompts. Extensive experiments on challenging benchmarks demonstrate that our method achieves strong performance, exhibiting superior accuracy, robustness to partial inputs, and a powerful capacity for novel class discovery.Our code and models are available at https://github.com/Tap12345/LangGS.