GS-Surrogate: Deformable Gaussian Splatting for Parameter Space Exploration of Ensemble Simulations
This work addresses the problem of efficiently exploring ensemble simulations for scientific domains, offering an incremental improvement over existing visualization surrogate models by providing an explicit 3D representation with separated variations.
The paper tackled the challenge of enabling flexible post-hoc exploration of ensemble simulations by introducing GS-Surrogate, a deformable Gaussian Splatting-based visualization surrogate that constructs a canonical Gaussian field and adapts it through parameter-conditioned deformations, resulting in real-time and controllable exploration across simulation and visualization parameter spaces.
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.