DualSplat：通过重建失败伪掩码自举的鲁棒三维高斯泼溅
DualSplat: Robust 3D Gaussian Splatting via Pseudo-Mask Bootstrapping from Reconstruction Failures
将首轮重建失败转化为显式先验，利用光度残差、特征不匹配和 SAM2 实例边界生成伪掩码，再引导第二轮干净重建，为处理训练图像中瞬态干扰提供可复用思路。
原文依据 We exploit these failures to construct object-level pseudo-masks by combining photometric residuals, feature mismatches, and SAM2 instance boundaries.
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论三维高斯泼溅（3DGS）虽能实现实时逼真渲染，但当训练图像包含违反多视图一致性的瞬态物体时，其性能会显著下降。现有方法面临循环依赖：准确的瞬态检测需要重建良好的静态场景，而干净的重建本身又依赖可靠的瞬态掩码。我们通过 DualSplat 应对这一挑战，这是一个“失败到先验”框架，将首轮重建失败转化为第二轮重建阶段的显式先验。我们观察到，仅出现在部分视图中的瞬态物体，在保守的初始训练中常表现为不完整碎片。我们利用这些失败，结合光度残差、特征不匹配和 SAM2 实例边界来构建物体级伪掩码。这些伪掩码随后引导干净的第二轮 3DGS 优化，同时一个轻量级 MLP 通过从先验监督逐渐转向自一致性来在线细化它们。在 RobustNeRF 和 NeRF On-the-go 上的实验表明，DualSplat 优于现有基线，在瞬态密集场景和瞬态区域中优势尤为明显。
英文原文摘要 1231 字符
While 3D Gaussian Splatting (3DGS) achieves real-time photorealistic rendering, its performance degrades significantly when training images contain transient objects that violate multi-view consistency. Existing methods face a circular dependency: accurate transient detection requires a well-reconstructed static scene, while clean reconstruction itself depends on reliable transient masks. We address this challenge with DualSplat, a Failure-to-Prior framework that converts first-pass reconstruction failures into explicit priors for a second reconstruction stage. We observe that transients, which appear in only a subset of views, often manifest as incomplete fragments during conservative initial training.We exploit these failures to construct object-level pseudo-masks by combining photometric residuals, feature mismatches, and SAM2 instance boundaries. These pseudo-masks then guide a clean second-pass 3DGS optimization, while a lightweight MLP refines them online by gradually shifting from prior supervision to self-consistency. Experiments on RobustNeRF and NeRF On-the-go show that DualSplat outperforms existing baselines, demonstrating particularly clear advantages in transient-heavy scenes and transient regions.