十年之后：利用智能体训练数据将阴影去除带入现实世界

中文摘要 AI 辅助

阴影去除在既有基准上看似已接近解决，但在现实世界中仍然脆弱。模型已经进步，但它们依赖的成对训练数据在近十年中几乎没有变化。原因很简单：获得无阴影的目标图像需要移除遮挡物，同时保持场景、相机和光照不变，这使得多样化的成对数据难以捕获。与此同时，大型阴影检测数据集已经包含多样化的现实世界图像和掩码，但没有无阴影的目标。为了将这些丰富但不完整的数据转化为成对监督，我们提出了一种离线智能体工作流，结合了物理驱动的生成、失败检测、反馈驱动的重试、候选选择和确定性校正。利用该工作流，我们构建了AgenticShadow数据集，包含17,138个图像-掩码-目标三元组，涵盖一般场景、人脸和遥感。我们的构建工作流将先前阴影去除工作的颜色分布差异降低了50.5%，而在AgenticShadow上训练现有阴影去除模型，将跨域LAB RMSE降低了19.7%至37.5%。

英文摘要

Shadow removal looks nearly solved on established benchmarks, yet remains brittle in the real world. Models have advanced; the paired training data they rely on have barely changed in nearly a decade. The reason is simple: obtaining a shadow-free target requires removing the occluder while keeping the scene, camera, and illumination otherwise unchanged, making diverse paired data difficult to capture. Meanwhile, large shadow detection datasets already contain diverse real-world images and masks, but no shadow-free targets. To turn this abundant but incomplete data into paired supervision, we propose an offline agentic workflow combining physics-motivated generation, failure detection, feedback-driven retry, candidate selection, and deterministic correction. Using this workflow, we construct AgenticShadow, a dataset of 17,138 image-mask-target triplets spanning general scenes, faces, and remote sensing. Our construction workflow reduces Color Distribution Difference by 50.5% over previous shadow removal work, while training existing shadow removal models on AgenticShadow reduces cross-domain LAB RMSE by 19.7-37.5%.

发表机构

• Stony Brook University（石溪大学）
• University of North Carolina at Charlotte（北卡罗来纳大学夏洛特分校）

机构由 AI 辅助整理，请以论文原文为准。