SceneLCM：基于 Latent Consistency Model 的端到端布局引导交互式室内场景生成
SceneLCM: End-to-End Layout-Guided Interactive Indoor Scene Generation with Latent Consistency Model
AI总结：
本文提出端到端框架 SceneLCM，结合 LLM 布局推理与 LCM 场景优化，实现布局、家具、环境和物理编辑全流程的高质量交互式室内场景生成。
摘要：
我们的项目页面：https://scutyklin.github.io/SceneLCM/。自动生成根据用户提示定制的复杂、可交互室内场景仍是一项艰巨挑战。现有方法虽能实现室内场景合成，却受限于僵化的编辑约束、物理不一致、过多人工投入、仅支持单房间以及材质质量欠佳等问题。为克服这些局限，我们提出 SceneLCM，这是一个端到端框架，将用于布局设计的 Large Language Model（LLM，大语言模型）与用于场景优化的 Latent Consistency Model（LCM，潜在一致性模型）协同结合。我们的方法把场景生成分解为四个模块化流程：（1）布局生成。我们采用 LLM 引导的三维空间推理，将文本描述转换为参数化蓝图（三维布局）。同时，迭代式程序化验证机制通过 LLM 介导的对话循环迭代细化布局参数；（2）家具生成。SceneLCM 采用 Consistency Trajectory Sampling（CTS，一致性轨迹采样），这是一种由 LCM 引导的一致性蒸馏采样损失，用于形成快速、语义丰富且高质量的表示。我们还给出两项理论论证，证明 CTS 损失等价于一致性损失，并且其蒸馏误差受 Euler solver 截断误差的约束；（3）环境优化。我们使用多分辨率纹理场编码场景外观，并通过 CTS 损失进行优化。为保持跨几何纹理一致性，我们引入法线感知交叉注意力解码器，通过交叉关注几何异质实例中的锚点位置来预测 RGB；（4）物理编辑。SceneLCM 通过集成物理仿真支持物理编辑，实现持久的物理真实感。大量实验验证了 SceneLCM 相对于最先进技术的优越性，显示其在多种应用中的广泛潜力。
英文摘要：
Our project page: https://scutyklin.github.io/SceneLCM/. Automated generation of complex, interactive indoor scenes tailored to user prompt remains a formidable challenge. While existing methods achieve indoor scene synthesis, they struggle with rigid editing constraints, physical incoherence, excessive human effort, single-room limitations, and suboptimal material quality. To address these limitations, we propose SceneLCM, an end-to-end framework that synergizes Large Language Model (LLM) for layout design with Latent Consistency Model(LCM) for scene optimization. Our approach decomposes scene generation into four modular pipelines: (1) Layout Generation. We employ LLM-guided 3D spatial reasoning to convert textual descriptions into parametric blueprints(3D layout). And an iterative programmatic validation mechanism iteratively refines layout parameters through LLM-mediated dialogue loops; (2) Furniture Generation. SceneLCM employs Consistency Trajectory Sampling(CTS), a consistency distillation sampling loss guided by LCM, to form fast, semantically rich, and high-quality representations. We also offer two theoretical justification to demonstrate that our CTS loss is equivalent to consistency loss and its distillation error is bounded by the truncation error of the Euler solver; (3) Environment Optimization. We use a multiresolution texture field to encode the appearance of the scene, and optimize via CTS loss. To maintain cross-geometric texture coherence, we introduce a normal-aware cross-attention decoder to predict RGB by cross-attending to the anchors locations in geometrically heterogeneous instance. (4)Physically Editing. SceneLCM supports physically editing by integrating physical simulation, achieved persistent physical realism. Extensive experiments validate SceneLCM's superiority over state-of-the-art techniques, showing its wide-ranging potential for diverse applications.
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。
