Scene Experts: Specializing in 3D Gaussian Splatting with Adaptive Decomposition
Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (5) · 4040-4048
DOI: 10.1609/aaai.v40i5.37407 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Xiaowen Fu und 7 weiteren Autor:innen.
Abstract (Originalsprache)
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's total capacity while maintaining comparable inference cost via sparse activation. To effectively train this high-capacity model, we propose two key innovations: (1) A progressive curriculum learning strategy that first trains all experts on a robust baseline before encouraging them to specialize on different scene components. (2) A novel opacity-aware regularization that penalizes inactive neural Gaussians, ensuring the expanded capacity is efficiently used. Extensive experiments demonstrate that MoE-GS substantially outperforms state-of-the-art methods on diverse benchmarks, significantly improving reconstruction fidelity while requiring a smaller or comparable Gaussian model size.
Zitieren
APA
Fu, X., Zhang, Y., Tang, Y., Zhang, H., Zhao, T., Guo, Y., Huang, Y., & Wang, J. (2026). Scene Experts: Specializing in 3D Gaussian Splatting with Adaptive Decomposition. Proceedings of the AAAI Conference on Artificial Intelligence, 40(5), 4040-4048. https://doi.org/10.1609/aaai.v40i5.37407
Vancouver
Fu X, Zhang Y, Tang Y, Zhang H, Zhao T, Guo Y, et al. Scene Experts: Specializing in 3D Gaussian Splatting with Adaptive Decomposition. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(5):4040-4048. doi:10.1609/aaai.v40i5.37407
BibTeX
@article{Fu2026,
author = {Fu, Xiaowen and Zhang, Yang and Tang, Yuhan and Zhang, Huazhong and Zhao, Tianxing and Guo, Yuhang and Huang, Yu and Wang, Jinbao},
title = {Scene Experts: Specializing in 3D Gaussian Splatting with Adaptive Decomposition},
journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
year = {2026},
volume = {40},
number = {5},
pages = {4040--4048},
doi = {10.1609/aaai.v40i5.37407}
}
EUVIMED