Scene Experts: Specializing in 3D Gaussian Splatting with Adaptive Decomposition
Xiaowen Fu, Yang Zhang, Yuhan Tang, Huazhong Zhang, Tianxing Zhao, Yuhang Guo, Yu Huang, Jinbao Wang
Abstract
Anchor-based 3D Gaussian Splatting (GS), exemplified by Scaffold-GS, achieves remarkable storage efficiency through a hybrid explicit-implicit representation. However, their reliance on a single, monolithic network to decode anchor features imposes a severe bottleneck on model capacity, often resulting in blurred details and view-dependent artifacts in complex scenes. To break this bottleneck, we introduce the concept of Scene Experts: a strategy that decomposes the task of modeling a complex scene across a collection of specialized sub-models. To realize the paradigm, we propose MoE-GS. Our approach designs the decoder as a Sparsely-Gated Mixture of Experts (MoE), which dramatically increases the model
BibTeX
@inproceedings{aaai2026_sceneexpertsspec,
title = {Scene Experts: Specializing in 3D Gaussian Splatting with Adaptive Decomposition},
author = {Xiaowen Fu and Yang Zhang and Yuhan Tang and Huazhong Zhang and Tianxing Zhao and Yuhang Guo and Yu Huang and Jinbao Wang},
booktitle = {AAAI 2026},
year = {2026}
}