Pair2Scene: Learning Local Object Relations for Procedural Scene Generation

[论文] Pair2Scene: Learning Local Object Relations for Procedural Scene Generation ## 论文概要 **研究领域**: cs.CV **作者**: Xingjian Ran, Shujie Zhang, Weipeng Zhong, Li Luo, Bo Dai **发布时间**: 2026-04-13 **arXiv**: [2604.11808](https://arxiv.org/abs/2604.11808) ## 中文摘要 生成高保真3D室内场景面临数据稀缺和复杂空间关系建模的挑战。本文提出Pair2Scene，一种基于局部对象关系学习的程序化场景生成框架。核心洞察是：对象放置主要依赖局部依赖关系而非全局分布。框架捕获两种对象间关系：遵循物理层次的支持关系和反映语义链接的功能关系。实验表明该方法能生成超出训练数据分布的复杂环境，同时保持物理和语义合理性。 ## 原文摘要 Generating high-fidelity 3D indoor scenes remains a significant challenge due to data scarcity and the complexity of modeling intricate spatial relations. Current methods often struggle to scale beyond training distribution to dense scenes or rely on LLMs/VLMs that lack the ability for precise spatial reasoning. Building on top of the observation that object placement relies mainly on local dependencies instead of information-redundant global distributions, in this paper, we propose Pair2Scene, a novel procedural generation framework that integrates learned local rules with scene hierarchies and physics-based algorithms. --- *自动采集于 2026-04-15* #论文 #arXiv #AI #小凯