BRDFusion: Physics Meets Generation for Urban Scene Inverse Rendering
Summary
BRDFusion combines physical modeling with generative priors to achieve high-quality inverse and forward rendering of urban scenes, enabling applications like novel-view relighting and dynamic object insertion.
View Cached Full Text
Cached at: 06/16/26, 11:34 AM
Paper page - BRDFusion: Physics Meets Generation for Urban Scene Inverse Rendering
Source: https://huggingface.co/papers/2606.17049
Abstract
BRDFusion combines physical modeling and generative priors to achieve high-quality inverse and forward rendering of urban scenes with precise control and artifact reduction.
Inverse renderingof urban scenes from captured videos enables numerous applications, including content creation and autonomous driving simulation. Physically-based rendering methods follow and control lighting physics, but suffer from reconstruction and rendering artifacts. Whilegenerative modelsproduce realistic videos, they offer limited consistency and controllability. We presentBRDFusion, a unified framework that combines two complementary models for inverse andforward rendering. Specifically,BRDFusionrecovers explicit, consistentscene propertieswithphysical modelingand alleviatesoptimization ambiguitywith generative priors. Duringforward rendering, the physical model providescontrollable renderingfrom the scene configuration, and the generative model denoises and fixes artifacts. Therefore, our method produces high-quality videos while allowing precise control, outperforming baselines in real and synthetic scenes. Moreover,BRDFusionsupportsnovel-view relighting,night simulation, anddynamic object insertion/editing. Project page: https://shigon255.github.io/brdfusion-page/
View arXiv pageView PDFProject pageGitHub10Add to collection
Models citing this paper1
#### Shigon/BRDFusion_checkpoints Updatedabout 8 hours ago
Datasets citing this paper2
#### Shigon/BRDFusion_data Viewer• Updatedabout 8 hours ago • 112k • 83 #### Shigon/BRDFusion_videos Viewer• Updatedabout 8 hours ago • 495 • 24
Spaces citing this paper0
No Space linking this paper
Cite arxiv.org/abs/2606.17049 in a Space README.md to link it from this page.
Collections including this paper0
No Collection including this paper
Add this paper to acollectionto link it from this page.
Similar Articles
SynCity 3000: Bootstrapping Scene-Scale 3D Diffusion
SynCity 3000 introduces a framework for generating large, globally coherent 3D scenes by adapting image-to-3D generators as convolutional operators, fine-tuned on synthetic scene data from a new data engine.
InfiniteDiffusion: Bridging Learned Fidelity and Procedural Utility for Open-World Terrain Generation
Introduces InfiniteDiffusion, a training-free algorithm that enables diffusion models to generate unbounded, seed-consistent terrain with constant-time random access, bridging learned fidelity and procedural utility. The Terrain Diffusion framework demonstrates interactive-rate realistic terrain generation with Earth-scale dynamic ranges.
GenRecon: Bridging Generative Priors for Multi-View 3D Scene Reconstruction
GenRecon introduces a method for 3D scene reconstruction that integrates generative 3D priors with multi-view image conditioning, achieving high-fidelity, editable mesh reconstructions of indoor environments and outperforming existing methods by 16%.
From Synthesis to Removal: Physics-Grounded Reflection Simulation and Diffusion-Based Video Dereflection
This paper presents a closed-loop framework combining physics-grounded reflection simulation, a diffusion-based video dereflection model (S2R-Removal), and a new benchmark (S2R-Bench), achieving state-of-the-art video reflection removal with fast inference.
PhysiFormer: Learning to Simulate Mechanics in World Space
PhysiFormer uses coordinate-space diffusion to generate physically-plausible 3D object motions without explicit inductive biases, enabling efficient multi-object reasoning and generalization to complex materials and geometries.