GHPT: Real-Time Relightable Gaussian Splatting using Hybrid Path Tracing
Abstract
3D Gaussian splatting (3DGS) has emerged as a promising approach for high-fidelity 3D scene representation. However,relighting and composition of Gaussian splatting remain challenging because path tracing is not directly applicable. Existing relighting methods for Gaussian splatting typically adopt either approximate rendering formulations or rely on Gaussian ray tracing, yielding low relighting performance and low rendering efficiency. To address these limitations,we propose Gaussian hybrid path tracing (GHPT), a three stage framework to acquire relightable Gaussian splatting models. The first stage utilizes planar-based Gaussian splatting reconstruction representation (PGSR) to enable multi-view consistent depth rendering and reconstruct the surface mesh of a scene. The second stage performs physically-based differentiable rendering on the obtained mesh to reconstruct the material maps and the environment map. The third stage utilizes factorized inverse path tracing (FIPT) on the Gbuffer rendered by the PGSR, and visibility and indirect illu mination are evaluated by hardware-accelerated ray tracing on the mesh with the material maps and the environment map reconstructed in the second stage. Experiments demonstrate that the relighting performance of GHPT outperforms the baselines, and our method can perform real-time relighting and composition of Gaussian splatting.
Our three-stage GS-based inverse rendering pipeline using hybrid path tracing. In the first stage, the scene geometry is reconstructed from multi-view images through planar-based Gaussians guided by normal priors. The second stage employs physically-based differentiable rendering to recover material properties of the mesh and the environment map. In the final stage, we use FIPT to recover material property of each Gaussian and take into account visibility and indirect illumination by ray tracing the underlying mesh.
Qualitative comparison of relighting results under different environment maps on the Synthetic4Relight and TensoIR Synthetic datasets.
Qualitative comparison of relighting results under different environment maps on Synthetic4Relight dataset.
Qualitative comparison of relighting results under different environment maps on TensoIR Synthetic dataset.
Quantitative comparison on the Synthetic4Relight and TensoIR Synthetic datasets.
Real-time relighting and composition results in the Garden, Kitchen, and Room scenes. From left to right: results without path tracing, results of our method, and results of our method relit by two environment maps, Bridge and Fireplace. Our method can produce more realistic results due to the presence of shadows and indirect illumination.
Video Presentation
BibTeX
@article{Bo2026GHPT,
title={GHPT: Real-Time Relightable Gaussian Splatting using Hybrid Path Tracing},
author={Bo, Jinyang and Dou, Fan and Quan, Wenrui and Liu, Shangxun and Xu, Yang and Zhang, Yuhe and Li, Kang and Geng,
Guohua},
Conference={IEEE/CVF Conference on Computer Vision and Pattern Recognition},
year={2026}
}