Gaussian Splashing: Unified Particles for Versatile Motion Synthesis and Rendering

1University of Utah, 2Zhejiang University, 3UCLA
*Equal Contributions

Gaussian Splashing (GSP) is a unified framework combining 3D Gaussian Splatting (3DGS) and Position-Based Dynamics (PBD).

Abstract

We demonstrate the feasibility of integrating physics-based animations of solids and fluids with 3D Gaussian Splatting (3DGS) to create novel effects in virtual scenes reconstructed using 3DGS. Leveraging the coherence of the Gaussian Splatting and Position-Based Dynamics (PBD) in the underlying representation, we manage rendering, view synthesis, and the dynamics of solids and fluids in a cohesive manner. Similar to GaussianShader, we enhance each Gaussian kernel with an added normal, aligning the kernel's orientation with the surface normal to refine the PBD simulation. This approach effectively eliminates spiky noises that arise from rotational deformation in solids. It also allows us to integrate physically based rendering to augment the dynamic surface reflections on fluids. Consequently, our framework is capable of realistically reproducing surface highlights on dynamic fluids and facilitating interactions between scene objects and fluids from new views.

Pipeline

The input to our system comprises multi-view images that capture a 3D scene. During the preprocessing stage, foreground objects are isolated and reconstructed. This is followed by point sampling to facilitate scene discretization for PBD simulation and Gaussian rendering. We train the Gaussian kernels using differentiable 3DGS, which takes into account appearance materials and lighting conditions. These kernels are animated using PBD, in conjunction with fluid particles, to tackle the dynamics of both solids and fluids within the scene. Finally, the dynamic scene is rendered into images. This rendering process includes detailed modeling of specular reflections, thereby providing visually accurate representations of the simulated interactions between solids and fluids.

Shadow Mapping

we demonstrate that GSP, with the addition of nearly soft shadows, significantly improves visual realism compared to vanilla 3DGS. Without these shadows, objects appear like flat layers pasted onto the background, lacking a sense of depth.

w/o shadow
w/ shadow

PBR Specular Modeling

we compare fluids rendered using purely diffuse materials with those that incorporated specular reflections. The fluids without specular reflection appear almost smoke-like, while the inclusion of specular term significantly enhances the realism of the fluids.

w/o specular
w/ specular

Inpainting

Displaced object exposes unseen areas that were originally covered to the camera. Since they are not present in the input image, 3DGS is unable to recover the color and texture information of these areas, leading to black smudges and dirty textures in the result. GSP remedies this issue using an inpainting trick.

w/o inpainting
w/ inpainting

Anisotropy Regularization

3DGS obtained from a static scene produces low-quality renders when Gaussian kernels undergo large rotational deformations. The anisotropy regularization is effective against this limitation.

w/o aniso.
w/ aniso.

BibTeX

@article{feng2024splashing,
  title={Gaussian Splashing: Unified Particles for Versatile Motion Synthesis and Rendering},
  author={Feng, Yutao and Feng, Xiang and Shang, Yintong and Jiang, Ying and Yu, Chang and Zong, Zeshun and Shao, Tianjia and Wu, Hongzhi and Zhou, Kun and Jiang, Chenfanfu and Yang, Yin},
  journal={arXiv preprint arXiv:2401.15318},
  year={2024}
}