GS-Surrogate: Deformable Gaussian Splatting for Parameter Space Exploration of Ensemble Simulations

2026-04-07

AI Summary

This work addresses the high storage overhead and low interactivity commonly encountered in posterior visualization of ensemble simulation data by proposing an explicit 3D proxy model based on deformable Gaussian splatting. For the first time, deformable Gaussian splatting is introduced into ensemble simulation visualization, leveraging a parameter-conditioned deformation network to efficiently adapt a canonical Gaussian field to diverse visualization tasks. This approach decouples simulation variability from visualization settings, enabling flexible and efficient exploration. Experimental results demonstrate that the method supports real-time, joint parameter exploration across multiple simulation datasets, significantly reducing storage requirements while substantially improving interactive performance and visual fidelity.

Abstract

Exploring ensemble simulations is increasingly important across many scientific domains. However, supporting flexible post-hoc exploration remains challenging due to the trade-off between storing the expensive raw data and flexibly adjusting visualization settings. Existing visualization surrogate models have improved this workflow, but they either operate in image space without an explicit 3D representation or rely on neural radiance fields that are computationally expensive for interactive exploration and encode all parameter-driven variations within a single implicit field. In this work, we introduce GS-Surrogate, a deformable Gaussian Splatting-based visualization surrogate for parameter-space exploration. Our method first constructs a canonical Gaussian field as a base 3D representation and adapts it through sequential parameter-conditioned deformations. By separating simulation-related variations from visualization-specific changes, this explicit formulation enables efficient and controllable adaptation to different visualization tasks, such as isosurface extraction and transfer function editing. We evaluate our framework on a range of simulation datasets, demonstrating that GS-Surrogate enables real-time and flexible exploration across both simulation and visualization parameter spaces.

Problem

• ensemble simulations
• parameter space exploration
• visualization surrogate
• interactive visualization
• 3D representation

Innovation

• Deformable Gaussian Splatting
• Visualization Surrogate
• Parameter Space Exploration
• Ensemble Simulations
• Explicit 3D Representation