Text-Driven 3D Indoor Scene Synthesis in Non-Manhattan Environments

Source

Source name: arXiv
Source kind: paper
Authors / organization: Xianhui Meng, Zirui Song, Yuchen Zhang, Li Zhang, Yongxuan Lv, Xiuying Chen, Kun Wang, Yan Luo, Kai Chen, Hangjun Ye, Long Chen, Jun Liu, Xiaoshuai Hao
Venue: arXiv: cs.AI
Published: 2026-07-02
Collected: 2026-07-03
URL: http://arxiv.org/abs/2607.02407v1
PDF: https://arxiv.org/pdf/2607.02407v1
DOI: UNKNOWN

Plain-Language Brief

Researchers have developed a new method called SPG-Layout for creating realistic indoor scenes in complex, non-rectangular (non-Manhattan) environments using large language models. This method improves upon existing ones by capturing more accurate object layout patterns and reducing geometric violations through statistical priors and hierarchical layout strategies.

What the Document Says

Large Language Models have shown capabilities in 3D indoor synthesis for Manhattan environments, but fail to capture plausible object layout patterns in non-Manhattan settings due to issues with modeling non-orthogonal spatial relationships.
To address this challenge, the authors propose SPG-Layout: a text-driven framework for generating physically plausible indoor scenes within complex non-Manhattan environments.
The framework utilizes statistical priors of object distributions to guide training and prioritize large object placement to minimize layout violations.
SPG-Layout is designed to achieve a balance between semantic realism and physical plausibility.
A new benchmark with 500 diverse non-Manhattan environments was created for evaluation. Experiments demonstrate SPG-Layout's consistent and significant performance improvement over existing methods in both Manhattan and non-Manhattan environments.
The authors plan to release the code publicly.

Key Concepts Mentioned

Large Language Models
3D indoor synthesis
non-Manhattan environments
SPG-Layout
statistical priors
object distributions
hierarchical layout strategy
large object placement
layout violations
semantic realism
physical plausibility
benchmark

Claims Made by the Document

Existing methods fail to capture plausible object layout patterns in non-Manhattan settings.
SPG-Layout generates physically plausible indoor scenes within complex non-Manhattan environments.
SPG-Layout uses statistical priors of object distributions to guide training and minimize layout violations.
SPG-Layout achieves a balanced optimization of semantic realism and physical plausibility.
Extensive experiments demonstrate SPG-Layout's consistent and significant performance improvement over existing methods.

Evidence Used

The text states that existing methods often fail to capture plausible object layout patterns in non-Manhattan settings, and the authors propose SPG-Layout as a solution.
The authors describe how SPG-Layout uses statistical priors of object distributions to guide training and prioritize large object placement.
The text mentions that SPG-Layout achieves a balanced optimization of semantic realism and physical plausibility.
The authors report extensive experiments demonstrating SPG-Layout's consistent and significant performance improvement over existing methods.

Methods / Approach

UNKNOWN

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

Large Language Models (LLMs) have demonstrated remarkable capabilities in 3D indoor synthesis for Manhattan environments. However, existing methods often fail to capture plausible object layout patterns in non-Manhattan settings, primarily because they struggle to model non-orthogonal spatial relationships, leading to high geometric violations and low physical fidelity. To address this challenge, we propose SPG-Layout, a novel text-driven framework designed to generate physically plausible indoor scenes within complex non-Manhattan environments. Specifically, we first utilize statistical priors of object distributions to guide the training process, enhancing environmental understanding and fidelity. Furthermore, mirroring human design workflows, we adopt a hierarchical layout strategy that prioritizes the placement of large objects, thereby substantially minimizing layout violations. By synergizing these components, SPG-Layout achieves a balanced optimization of semantic realism and physical plausibility. To evaluate performance in these complex settings, we constructed a new benchmark comprising 500 diverse non-Manhattan environments. Extensive experiments demonstrate that SPG-Layout consistently and significantly outperforms existing methods across both Manhattan and non-Manhattan environments. The code will be publicly released.