Bridge：基驱动因果推理结合视觉基础模型实现域泛化
Bridge: Basis-Driven Causal Inference Marries VFMs for Domain Generalization
把因果前门调整实现为可学习的低秩基，用来阻断光照、风格、共现等混杂因子对检测器的虚假关联，并可作为即插即用模块挂到 DINOv2/3、SAM、Stable Diffusion 等视觉基础模型上，为单源域泛化检测提供了一条不依赖目标域数据的通用思路。
原文依据 By learning the low-rank bases for front-door adjustment, Bridge blocks confounders' effects to mitigate spurious correlations
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论检测器常常出现性能下降，主要原因是源域与目标域之间的分布差异。这一问题在数据有限的单源域场景中尤为明显，因为模型倾向于依赖源域中的混杂因子（如光照、共现和风格），从而产生阻碍泛化的虚假关联。为此，本文提出了一种新的面向域泛化的基驱动框架 Bridge，将因果推理引入目标检测。通过学习用于前门调整的低秩基，Bridge 阻断混杂因子的影响以缓解虚假关联，同时通过过滤冗余且与任务无关的成分来精炼表征。Bridge 可以无缝集成到判别式（如 DINOv2/3、SAM）和生成式（如 Stable Diffusion）视觉基础模型（VFM）中。在多个域泛化目标检测数据集上的大量实验，即 Cross-Camera、Adverse Weather、Real-to-Artistic、Diverse Weather Datasets 以及 Diverse Weather DroneVehicle（我们新扩充的基于真实世界无人机基准），凸显了所提方法相较于此前最优方法的优越性。项目页面见：https://mingbohong.github.io/Bridge/。
英文原文摘要 1343 字符
Detectors often suffer from degraded performance, primarily due to the distributional gap between the source and target domains. This issue is especially evident in single-source domains with limited data, as models tend to rely on confounders (e.g., illumination, co-occurrence, and style) from the source domain, leading to spurious correlations that hinder generalization. To this end, this paper proposes a novel Basis-driven framework for domain generalization, namely Bridge, that incorporates causal inference into object detection. By learning the low-rank bases for front-door adjustment, Bridge blocks confounders' effects to mitigate spurious correlations, while simultaneously refining representations by filtering redundant and task-irrelevant components.Bridge can be seamlessly integrated with both discriminative (e.g., DINOv2/3, SAM) and generative (e.g., Stable Diffusion) Vision Foundation Models (VFMs). Extensive experiments across multiple domain generalization object detection datasets, i.e., Cross-Camera, Adverse Weather, Real-to-Artistic, Diverse Weather Datasets, and Diverse Weather DroneVehicle (our newly augmented real-world UAV-based benchmark), underscore the superiority of our proposed method over previous state-of-the-art approaches. The project page is available at: https://mingbohong.github.io/Bridge/.