为稀疏视角高斯泼溅丢弃锚点与球谐函数
Dropping Anchor and Spherical Harmonics for Sparse-view Gaussian Splatting
针对稀疏视角 3DGS 过拟合，提出以锚点为中心连带删除空间邻居的 dropout，并随机丢弃高阶球谐系数；该策略可直接插入现有 3DGS 变体作为正则化手段，且 SH 截断可用于训练后压缩。
原文依据 we propose DropAnSH-GS, a novel anchor-based dropout strategy
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论近期的三维高斯泼溅（3DGS）dropout 方法通过随机将高斯不透明度置零来缓解稀疏视角条件下的过拟合。然而，我们发现这些方法中存在邻居补偿效应：被丢弃的高斯往往由其邻居补偿，从而削弱了原本期望的正则化效果。此外，这些方法忽略了高阶球谐系数（SH）对过拟合的影响。为解决上述问题，我们提出 DropAnSH-GS，一种新颖的基于锚点的 dropout 策略。该方法不再独立地丢弃高斯，而是随机选取部分高斯作为锚点，并同时移除其空间邻居。这有效破坏了局部冗余，促使模型学习更鲁棒、具备全局信息的表征。进一步地，我们将 dropout 扩展到颜色属性，通过随机丢弃高阶球谐系数，使外观信息集中于低阶球谐系数。该策略进一步缓解过拟合，并可通过球谐截断实现灵活的训练后模型压缩。实验结果表明，DropAnSH-GS 以可忽略的计算开销显著优于现有 dropout 方法，且可便捷地集成到各类 3DGS 变体中提升其性能。
英文原文摘要 1311 字符
Recent 3D Gaussian Splatting (3DGS) dropout methods address overfitting under sparse-view conditions by randomly nullifying Gaussian opacities. However, we identify a neighbor compensation effect in these approaches: dropped Gaussians are often compensated by their neighbors, weakening the intended regularization. Moreover, these methods overlook the contribution of high-degree spherical harmonic coefficients (SH) to overfitting. To address these issues, we propose DropAnSH-GS, a novel anchor-based dropout strategy. Rather than dropping Gaussians independently, our method randomly selects certain Gaussians as anchors and simultaneously removes their spatial neighbors. This effectively disrupts local redundancies and encourages the model to learn more robust, globally informed representations. Furthermore, we extend the dropout to color attributes by randomly dropping higher-degree SH coefficients to concentrate appearance information in lower-degree SH. This strategy further mitigates overfitting and enables flexible post-training model compression via SH truncation. Experimental results demonstrate that DropAnSH-GS substantially outperforms existing dropout methods with negligible computational overhead and can be readily integrated into various 3DGS variants to enhance their performances.