建模人类草图的视觉歧义性
Modeling the Visual Ambiguity of Human Sketches
AmbiScore 把「草图-图像对本身有多歧义」量化成一个可计算的指标，并据此自适应调节监督强度；做跨模态对齐或噪声监督训练时，可以直接借用这种按样本不确定性加权监督的思路，Purified Matching 的 shape jigsaw / texture swapping 也可作为解耦结构与外观的通用数据构造手段。
原文依据 we introduce AmbiScore, a metric that quantifies the ambiguity of sketch-image pairs
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论人类草图是一种紧凑而富有表现力的视觉交流形式，但其稀疏的结构线索在捕捉物体本质结构的同时也引入了歧义：同一张草图可以对应多张合理的图像，从而使跨域对齐变得不确定且不稳定。这种歧义从根本上限制了依赖精确草图—图像对应关系的草图视觉任务。为应对这一挑战，我们提出 AmbiScore，一种量化草图—图像对歧义程度的指标，并以零样本草图图像检索（ZS-SBIR）作为试验平台，揭示歧义性监督如何导致现有方法性能崩溃。我们进一步提出 DisAmb（解耦歧义）框架，通过两个组件显式建模并缓解歧义：（1）弹性匹配，利用 AmbiScore 自适应调整监督强度；（2）纯化匹配，借助与歧义无关的掩码，通过形状拼图与纹理交换解耦结构与外观。DisAmb 在高歧义条件下建立了新的基准，并为下游草图引导任务提供了稳健、可迁移的监督信号。
英文原文摘要 1194 字符
Human sketches provide a compact and expressive form of visual communication, but their sparse structural cues, while capturing essential object structures, introduce ambiguity because a single sketch can correspond to multiple plausible images, making cross-domain alignment uncertain and unstable. Such ambiguity fundamentally limits sketch-based vision tasks that rely on precise sketch--image correspondence. To address this challenge, we introduce AmbiScore, a metric that quantifies the ambiguity of sketch-image pairs, and use Zero-Shot Sketch-Based Image Retrieval (ZS-SBIR) as a testbed to reveal how ambiguous supervision leads to performance collapse in existing methods. We further propose DisAmb (Disentangling Ambiguity), a framework that explicitly models and mitigates ambiguity through two components: (1) Elastic Matching, which adaptively adjusts supervision strength using AmbiScore, and (2) Purified Matching, which employs ambiguity-agnostic masks to disentangle structure and appearance via shape jigsaw and texture swapping. DisAmb establishes new benchmarks under high ambiguity and provides a robust, transferable supervisory signal for downstream sketch-guided tasks.