SpotlessGS: Relightable 3D Gaussian Splatting under Dynamic Illumination for Robotic Perception

2026-08-11

AI Summary

This study addresses the degradation of robotic perception and multi-view geometric inconsistencies in low-light environments by extending the DarkGS framework with a relightable 3D Gaussian Splatting reconstruction method. By jointly optimizing illumination parameters and integrating low-frequency spherical harmonics with an MLP-based BRDF, the approach accurately models complex lighting and non-Lambertian reflectance without explicit calibration. This enables high-fidelity reconstruction under dynamic illumination conditions. Experimental results demonstrate that the proposed method effectively eliminates lighting artifacts and achieves rendering quality significantly superior to existing state-of-the-art techniques. Furthermore, it substantially enhances robustness in downstream robotic perception tasks, offering a practical solution for reliable visual sensing in challenging low-light scenarios.

Abstract

Robots operating in dark or poorly lit environments rely on onboard lights, which often produce uneven illumination that degrades downstream perception tasks. Prior approaches based on 2D image enhancement lack reliable supervision and fail to preserve multi-view geometric consistency. To address these limitations, we extend Dark Gaussian Splatting (DarkGS) toward a more accurate and flexible relightable 3D reconstruction framework. First, we eliminate the need for explicit light parameter calibration by jointly optimizing lighting parameters within the Gaussian Splatting framework. Second, we introduce a low-frequency illumination model based on spherical harmonics (SH) to capture spatially varying residual and ambient lighting effects. Third, we incorporate an MLP-based Bidirectional Reflectance Distribution Function (BRDF) to model non-Lambertian reflectance. Experiments on synthetic and real-world datasets demonstrate that our method effectively mitigates illumination artifacts while improving rendering quality and quantitative performance over prior approaches. We further validate its benefits for robotic perception through a downstream task.

Problem

• Robotic Perception
• Dynamic Illumination
• 3D Gaussian Splatting
• Relighting
• Uneven Illumination

Innovation

• Relightable 3D Gaussian Splatting
• Joint Lighting Optimization
• Spherical Harmonics Illumination Model
• MLP-based BRDF
• Robotic Perception