[论文] SimuScene: Simulation-Ready Compositional 3D Scene Reconstruction from...

论文概要

研究领域: CV
作者: Inhee Lee, Sangwon Baik, Sungjoo Kim, Hyeonwoo Kim, Hyunsoo Cha, Hanbyul Joo
发布时间: 2026-06-02
arXiv: 2606.03994

中文摘要

从单张图像重建可交互、可仿真的3D场景是机器人操作的关键瓶颈。虽然最近的单图像提升器能够恢复合理的物体形状，但将它们组合起来会产生在物理仿真中崩溃的场景，因为物体会相互穿透、悬浮或下沉。现有的物理感知方法严格将其作为事后布局修正，而未能解决底层几何误差。为此，我们引入了SimuScene，一种将物理纳入形状和布局估计循环的组合式3D重建管道。我们不是将物理仅用于布局清理，而是将物理引擎用作生成过程中的诊断测量工具。通过对重力下重建物体的诊断仿真，我们将穿透和支撑失败转化为驱动重力轴拉伸和非模态形状重采样的定量校正信号。这种物理知情反馈循环缓解了累积的重建误差，产生稳定的、可仿真就绪的组合式3D场景。大量实验证明了在物理稳定性和几何对齐基准上的最先进性能。

原文摘要

Reconstructing interactive, simulation-ready 3D scenes from a single image is a critical bottleneck for robotic manipulation. While recent single-image lifters recover plausible per-object shapes, composing them yields scenes that collapse under physical simulation due to interpenetrating, hovering, or sinking objects. Existing physics-aware methods address this strictly as a post-hoc layout correction, leaving the underlying geometric errors unresolved. To address this, we introduce SimuScene, a compositional 3D reconstruction pipeline that puts physics in the loop of shape and layout estimation. Rather than using physics merely for layout cleanup, we utilize the physics engine as a diagnostic measurement tool during the generative process itself. By diagnostically simulating reconstructe...

自动采集于 2026-06-04

#论文 #arXiv #CV #小凯