arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.03809cs.GRcs.CV
收藏

Split&Splat: 通过显式实例建模和3D高斯点撒技术实现零样本全景分割

Split&Splat: Zero-Shot Panoptic Segmentation via Explicit Instance Modeling and 3D Gaussian Splatting

University of Padova（帕多瓦大学）

Leonardo Monchieri, Elena Camuffo, Francesco Barbato, Pietro Zanuttigh, Simone Milani

阅读 PDF
arXiv 页面 ↗HTML 原文 ↗

AI总结：

Split&Splat通过显式实例建模和3D高斯点撒技术实现零样本全景分割，提升场景重建的精度和效率。

摘要：

3D高斯点撒（GS）能够实现快速且高质量的场景重建，但缺乏一致的对象和语义感知结构。我们提出Split&Splat，一种使用3DGS进行全景场景重建的框架。我们的方法显式地建模对象实例。首先利用深度将实例掩码传播到不同视角，从而生成视角一致的2D掩码。然后每个对象被独立重建并合并回场景中，同时细化其边界。最后，实例级语义描述符被嵌入到重建的对象中，支持各种应用，包括全景分割、对象检索和3D编辑。与现有方法不同，Split&Splat通过首先分割场景然后单独重建每个对象来解决该问题。这种设计自然支持下游任务，并使Split&Splat在ScanNetv2分割基准上实现了最先进的性能。

英文摘要：

3D Gaussian Splatting (GS) enables fast and high-quality scene reconstruction, but it lacks an object-consistent and semantically aware structure. We propose Split&Splat, a framework for panoptic scene reconstruction using 3DGS. Our approach explicitly models object instances. It first propagates instance masks across views using depth, thus producing view-consistent 2D masks. Each object is then reconstructed independently and merged back into the scene while refining its boundaries. Finally, instance-level semantic descriptors are embedded in the reconstructed objects, supporting various applications, including panoptic segmentation, object retrieval, and 3D editing. Unlike existing methods, Split&Splat tackles the problem by first segmenting the scene and then reconstructing each object individually. This design naturally supports downstream tasks and allows Split&Splat to achieve state-of-the-art performance on the ScanNetv2 segmentation benchmark.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。

↑