UrbanIR: Large-Scale Urban Scene Inverse Rendering from a Single Video
Chih-Hao Lin, 刘博涵 (Bohan Liu), Yi-Ting Chen, Kuan-Sheng Chen, David Forsyth, Jia-Bin Huang, Anand Bhattad, Shenlong Wang
摘要
此摘要由英文原文自动翻译。
我们提出UrbanIR(Urban Scene Inverse Rendering)——一个新的逆图形学模型,能从单段视频实现各种光照条件下场景的真实感自由视角渲染。它能从宽基线视频(如车载摄像机)精确推断形状、反照率、可见度以及太阳与天空光照,有别于NeRF的密集视角设定。在此情形下,标准方法常产生不佳的几何与材质估计,例如不精确的屋顶表征与大量「漂浮物」。UrbanIR以新损失函数解决这些问题,降低逆图形推断误差与渲染伪影。其技术能精确估计原始场景中的阴影体积。模型输出支持可控编辑,可实现夜间模拟、重新光照场景与插入物体的逼真自由视角渲染,标志着对现有最先进方法的显著改进。
原始摘要(英文)
We present UrbanIR (Urban Scene Inverse Rendering), a new inverse graphics model that enables realistic, free-viewpoint renderings of scenes under various lighting conditions with a single video. It accurately infers shape, albedo, visibility, and sun and sky illumination from wide-baseline videos, such as those from car-mounted cameras, differing from NeRF's dense view settings. In this context, standard methods often yield subpar geometry and material estimates, such as inaccurate roof representations and numerous 'floaters'. UrbanIR addresses these issues with novel losses that reduce errors in inverse graphics inference and rendering artifacts. Its techniques allow for precise shadow volume estimation in the original scene. The model's outputs support controllable editing, enabling photorealistic free-viewpoint renderings of night simulations, relit scenes, and inserted objects, marking a significant improvement over existing state-of-the-art methods.
本文贡献
从单段随手拍摄的视频进行大规模城市场景逆渲染:将城市分解为反照率、几何与瞬态光照,实现远超以往单张图像方法的逼真重光照与编辑。
BibTeX
@inproceedings{lin2025urbanir,
title = {UrbanIR: Large-Scale Urban Scene Inverse Rendering from a Single Video},
author = {Chih-Hao Lin and Bohan Liu and Yi-Ting Chen and Kuan-Sheng Chen and David Forsyth and Jia-Bin Huang and Anand Bhattad and Shenlong Wang},
booktitle = {International Conference on 3D Vision (3DV)},
year = {2025},
url = {https://arxiv.org/abs/2306.09349}
}