MoE-GS Studio is a research hub for Mixture-of-Experts (MoE) architectures for Dynamic Gaussian Splatting.
This repository organizes our research exploring how expert specialization and routing mechanisms can improve dynamic 3D scene reconstruction.
Our work investigates how multiple dynamic Gaussian splatting models can be combined through Mixture-of-Experts frameworks to better handle diverse motion patterns and scene dynamics.
MoDE (Mixture of Deformation Experts) is our latest research project in the MoE-GS Studio series, extending the Mixture-of-Experts paradigm for dynamic Gaussian splatting.
Instead of relying on a single deformation model, MoDE introduces multiple deformation experts, each specializing in different motion behaviors.
A routing mechanism dynamically selects or combines experts depending on the spatial and temporal characteristics of the scene.
This design enables the model to better handle:
- complex object motion
- non-rigid deformation
- heterogeneous dynamic regions
MoDE builds upon insights from our earlier work MoE-GS, which demonstrated the benefits of combining multiple dynamic Gaussian splatting models.
š§ Code Release
The official implementation of MoDE will be released in this repository after the paper is accepted.
MoE-GS is the first work in our MoE-based dynamic Gaussian splatting research line.
It introduces a Mixture-of-Experts framework that adaptively combines multiple dynamic Gaussian splatting models.
Different models exhibit strengths under different conditionsāfor example:
- some handle fast motion
- others reconstruct fine geometric details
MoE-GS learns a pixel-wise routing mechanism that dynamically selects the most suitable expert during rendering.
Paper: ICLR 2026
Project repository:
https://github.com/cvsp-lab/MoE-GS