[论文] OuroWorld: Bringing Any 3D World Alive as Diverse, Endlessly Looping 3...

论文概要

研究领域: CV
作者: You-Zhe Xie, Ting-Wei Chou, Yu-Hsuan Li, Kaipeng Zhang, Zhixiang Wang, Yu-Lun Liu
发布时间: 2026-10-08
arXiv: 2610.12461

中文摘要

近期 3D 世界模型能生成逼真、可探索的场景，但它们在时间上被冻结了。OuroWorld 是一个无需掩码的框架，将任何静态 3D 高斯泼溅场景转化为 3D 动态图（cinemagraph）：生动、多样、可从任意视角无缝循环播放的动态场景。视觉-语言模型推断合理的动态模式并引导视频模型合成参考视频，再将其提升并补全为多视角视频。为从不完美的监督中学习，我们提出不一致性鲁棒的周期性 4DGS：傅里叶级数变形场在构造上保证循环性，锚定在参考视角的接地漂移场（Grounded Drift Field）吸收跨视角不一致。与先前仅限流体类运动的欧拉方法不同，我们捕捉一般变形、物体运动和光照变化。评估无需真实标注，涵盖生动性、自然度、循环接缝连贯性和场景质量。在 39 个重建和生成场景上，OuroWorld 超越所有基线，用户研究中赢得 70.8%-99.0% 的对比。

原文摘要

Recent 3D world models generate photorealistic, explorable scenes that remain frozen in time. OuroWorld is a mask-free framework that turns any static 3D Gaussian Splatting scene into a 3D cinemagraph: a dynamic scene with vivid, diverse motion looping seamlessly from any viewpoint. A vision-language model infers plausible dynamics and guides a video model to synthesize a reference video, which we lift and complete into multi-view videos. To learn from this imperfect supervision, we propose Inconsistency-Robust Periodic 4DGS: a Fourier-series deformation field guarantees looping by construction, while a Grounded Drift Field anchored at the reference view absorbs cross-view inconsistency. Unlike prior Eulerian methods limited to fluid-like motion, we capture general deformation, object moti...

自动采集于 2026-10-10

#论文 #arXiv #CV #小凯