A Generalizable Light Transport 3D Embedding for Global Illumination
- Auto-discovered
Bing Xu, Mukund Varma T, Cheng Wang, Tzu-Mao Li, Lifan Wu, Bartlomiej Wronski, Ravi Ramamoorthi, Marco Salvi
Visit resource ↗Description
A Generalizable Light Transport 3D Embedding for Global Illumination is the SIGGRAPH 2026 paper by Bing Xu, Mukund Varma T, Cheng Wang, Tzu-Mao Li, Lifan Wu, Bartlomiej Wronski, Ravi Ramamoorthi, and Marco Salvi on cross-scene global illumination in 3D. The ACM Digital Library page is the canonical DOI record for the conference version.
Scenes become point clouds with material features. A linear-complexity transformer encodes long-range transport into neural primitives; query points gather k-nearest primitives with cross-attention to predict irradiance without rasterized or path-traced lighting cues.
Results cover diffuse GI on varied indoor scenes, fine-tuning for glossy radiance fields, and preliminary path-guiding warm starts. The model targets view-consistent predictions beyond prior object-centric neural GI limits.
Link last verified accessible on 2026-09-27.
Related resources
- Gabor Fields: Orientation-Selective Level-of-Detail for Volume Rendering
Gabor Fields: Orientation-Selective Level-of-Detail for Volume Rendering presents anisotropic Gabor kernels as an alternative to Gaussian primitives for volumetric path tracing,…
100% match · <code><math><advanced><paper><accessible><last><link><path> - Occlusion-Point Reuse for Ray-Traced Ambient Occlusion and Shadow
Occlusion-Point Reuse for Ray-Traced Ambient Occlusion and Shadow is a paper by Haojie Jin, Fujia Su, Zehui Lin, Chenxiao Hu, Jierui Ren, Yuqing Yuan, Yanchen Zhang, Zhongtao…
100% match · <code><advanced><paper><wang><accessible><last><light><link> - Meshtryoshka: Differentiable Mesh Rendering for Unbounded Scenes
Meshtryoshka: Differentiable Mesh Rendering for Unbounded Scenes introduces nested mesh shells extracted from a signed distance field and rasterized with alpha compositing for…
100% match · <code><math><advanced><paper><accessible><based><last><link> - Neural Render Proxies for Interactive and Differentiable Lighting
Neural Render Proxies for Interactive and Differentiable Lighting introduces a scene-specific neural render proxy that decouples path sampling from emission so static scenes with…
100% match · <code><advanced><paper><accessible><based><last><link><materials> - Illumination-Aware Spatial Subdivision for Path Guiding
Illumination-Aware Spatial Subdivision for Path Guiding proposes splitting path-guiding k-d trees using lookahead cells that store compact light-field signatures (mean radiance…
100% match · <code><math><advanced><paper><accessible><guiding><illumination><last> - Reconstructing Physically Stable 3D Scenes from a Single Image
Reconstructing Physically Stable 3D Scenes from a Single Image is a single research paper with a public REST3D project site, code release, and demo videos.
100% match · <code><math><advanced><paper><wang><accessible><geometry><last> - Dual Contouring of Signed Distance Data
Dual Contouring of Signed Distance Data is a SIGGRAPH 2026 paper with a Columbia Graphics project page linking the PDF, source, and talk materials.
100% match · <code><math><advanced><paper><wang><accessible><geometry><last> - Gaussian Point Splatting
Gaussian Point Splatting proposes a stochastic method for rendering massive 3D Gaussian Splatting scenes by sampling pixel-sized opaque points from Gaussians, splatting via…
100% match · <code><advanced><paper><accessible><based><last><link><point> - Precomputed Lens Transport Maps
Precomputed Lens Transport Maps is an ACM Transactions on Graphics paper by University of Waterloo authors, hosted on arXiv as a PDF.
100% match · <code><math><advanced><paper><accessible><geometry><last><link> - ReSTIR PT Enhanced: Algorithmic Advances for Faster and More Robust ReSTIR Path Tracing
ReSTIR PT Enhanced: Algorithmic Advances for Faster and More Robust ReSTIR Path Tracing presents algorithmic and engineering improvements to ReSTIR path tracing that report 2–3x…
100% match · <code><advanced><paper><accessible><based><global><illumination><last> - Parallelised Differentiable Straightest Geodesics for 3D Meshes
Parallelised Differentiable Straightest Geodesics for 3D Meshes presents a CIRCLE Group Imperial College London CVPR 2026 paper by Hippolyte Verninas, Caner Korkmaz, Stefanos…
100% match · <code><math><advanced><paper><accessible><last><link><transport> - MeshSplatting: Differentiable Rendering with Opaque Meshes
MeshSplatting: Differentiable Rendering with Opaque Meshes presents a mesh-based novel view synthesis method by Jan Held and collaborators that jointly optimizes geometry and…
100% match · <code><math><advanced><paper><accessible><based><geometry><last> - Sample Space Partitioning and Spatiotemporal Resampling for Specular Manifold Sampling
Sample Space Partitioning and Spatiotemporal Resampling for Specular Manifold Sampling presents a Utah Graphics Lab method by Pengpei Hong, Meng Duan, Beibei Wang, Cem Yuksel,…
100% match · <code><math><advanced><paper><wang><accessible><based><last> - Implicit Position-Based Fluids (IPBF)
Implicit Position-Based Fluids (IPBF) presents a Utah Graphics Lab incompressible SPH method by Elie Diaz, Jerry Hsu, Eisen Montalvo-Ruiz, Chris Giles, and Cem Yuksel that uses…
100% match · <code><math><advanced><paper><accessible><based><last><link> - Practically Utilizing Neural Networks in CPU-based Production Rendering
Practically Utilizing Neural Networks in CPU-based Production Rendering presents a Journal of Computer Graphics Techniques paper by Steve Bako, Ryusuke Villemin, and Magnus…
100% match · <code><math><advanced><paper><accessible><based><embedding><last> - OpenPBR: Novel Features and Implementation Details
OpenPBR: Novel Features and Implementation Details is a technical companion paper by Jamie Portsmouth (Autodesk), Peter Kutz (Adobe), and Stephen Hill (Lucasfilm) for the OpenPBR…
100% match · <math><advanced><paper><accessible><based><diffuse><last><light> - Optimizing spatiotemporal variance-guided filtering for modern GPU architectures
Optimizing spatiotemporal variance-guided filtering for modern GPU architectures.
100% match · <code><advanced><paper><accessible><based><last><link><path> - Nexels: Neurally-Textured Surfels for Real-Time Novel View Synthesis with Sparse Geometries
Nexels: Neurally-Textured Surfels for Real-Time Novel View Synthesis with Sparse Geometries presents a research project page by Victor Rong, Jan Held, Victor Chu, Daniel Rebain,…
100% match · <code><math><advanced><paper><accessible><based><geometry><global> - Iterative Lightmap Updates for Scene Editing
Iterative Lightmap Updates for Scene Editing is a Pacific Graphics 2025 conference paper by Guowei Lu, Christoph Peters, Petr Kellnhofer, and Elmar Eisemann, hosted on the…
100% match · <math><advanced><paper><accessible><global><illumination><last><link> - Jackknife Transmittance and MIS Weight Estimation
Jackknife Transmittance and MIS Weight Estimation is a SIGGRAPH Asia 2025 paper by Christoph Peters, published in ACM Transactions on Graphics and summarized on the Moments in…
100% match · <math><advanced><paper><accessible><last><link><path><point>