证据神经辐射场
Evidential Neural Radiance Fields
把证据深度学习（evidential deep learning）接进 NeRF 的渲染管线，用单次前向传播同时给出偶然不确定性与认知不确定性，可直接复用到任何需要逐像素置信度的三维重建/新视角合成流程中，用于安全关键场景的失效检测与主动视角选择。
原文依据 enabling direct quantification of both aleatoric and epistemic uncertainties from a single forward pass
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论理解不确定性来源是可信三维场景建模的基础。尽管神经辐射场（NeRF）的最新进展在场景重建与新视角合成上取得了令人印象深刻的精度，但缺乏不确定性估计严重限制了其在安全关键场景中的部署。现有的 NeRF 不确定性量化方法无法同时分别刻画偶然不确定性与认知不确定性。在那些只量化其中之一的方方法中，许多要么损害渲染质量，要么为获得不确定性估计而带来显著的计算开销。为解决这些问题，我们提出证据神经辐射场（Evidential Neural Radiance Fields），一种与 NeRF 渲染过程无缝集成的概率方法，能够从单次前向传播中直接量化偶然不确定性与认知不确定性。我们在三个标准化基准上比较了多种不确定性量化方法，我们的方法在场景重建保真度与不确定性估计质量上均达到最优。代码见 https://github.com/KerryDRX/EvidentialNeRF。
英文原文摘要 1166 字符
Understanding sources of uncertainty is fundamental to trustworthy three-dimensional scene modeling. While recent advances in neural radiance fields (NeRFs) achieve impressive accuracy in scene reconstruction and novel view synthesis, the lack of uncertainty estimation significantly limits their deployment in safety-critical settings. Existing uncertainty quantification methods for NeRFs fail to separately capture both aleatoric and epistemic uncertainties. Among those that do quantify one or the other, many of them either compromise rendering quality or incur significant computational overhead to obtain uncertainty estimates. To address these issues, we introduce Evidential Neural Radiance Fields, a probabilistic approach that seamlessly integrates with the NeRF rendering process, enabling direct quantification of both aleatoric and epistemic uncertainties from a single forward pass. We compare multiple uncertainty quantification methods on three standardized benchmarks, where our approach demonstrates state-of-the-art scene reconstruction fidelity and uncertainty estimation quality. Code is available at https://github.com/KerryDRX/EvidentialNeRF.