论文

CoST:用补充语义指导改进单正例多标签学习

Seeing Beyond the Obvious: Complementary Semantic Guidance for Salient Single-Positive Multi-Label Learning

精选理由

搞多标签学习或弱标注训练的值得看看这篇:单正例标注下怎么补足缺失类别的监督信号,思路是用图像-文本语义从未知类里筛候选建 Coverage Prototypes。做科研或处理只有部分标签数据集的团队可以借鉴这套软目标精炼策略。

一篇 arXiv 论文提出 CoST,针对 Salient Single-Positive Multi-Label Learning(SalSPML)任务中每张图像仅标注一个最显著正例标签、其余类别未标注的问题。

正文

现有方法依赖观测正例构建类别级 Salient Prototypes,但显著标注偏向视觉突出实例,难以覆盖隐含正例的视觉多样性。CoST 引入 Coverage-guided Complementary Soft-Target Refinement 框架,利用任务学习到的类别信息和图像-文本语义对应,从未知类别中筛选语义相关候选,聚合为 Coverage Prototypes 来补充 Salient Prototypes。

论文还设计了基于排序的选择性精炼策略,只对从 Coverage Prototypes 获得更强语义支持的类别增强其软目标,其余保留基础软目标。在多个基准上的实验显示 CoST 相对现有基线取得稳定性能提升。

原文 · arXiv cs.LG

Seeing Beyond the Obvious: Complementary Semantic Guidance for Salient Single-Positive Multi-Label Learning

Salient Single-Positive Multi-Label Learning (SalSPML) assigns each image its most salient positive label, leaving other classes unlabeled. Existing methods construct class-level Salient Prototypes from observed positives to guide the recognition of unobserved classes. However, since salient annotations favor visually prominent instances, these prototypes may not fully capture the visual diversity of hidden positives, resulting in insufficient positive supervision. To address this limitation, we propose CoST, a Coverage-guided Complementary Soft-Target Refinement framework that supplements Salient Prototypes with complementary class-semantic information. Specifically, CoST identifies semantically relevant candidates from unknown classes by leveraging task-learned category information and image-text semantic correspondence. The selected candidates are then aggregated into class-wise Coverage Prototypes to complement the semantic representations provided by Salient Prototypes. Furthermore, we introduce a ranking-based selective refinement strategy that identifies classes receiving stronger semantic support from Coverage Prototypes and selectively enhances their soft targets while preserving the base soft targets for the rest. Experiments on multiple benchmarks demonstrate the effectiveness of CoST, with strong performance gains over existing baselines.

产业坐标与信源核验

本条情报归属 AITOP「论文」核心追踪栏目。在跨区域产业观察坐标中,该动态直接关联「AI 开发者工具链、Coding Agent 与工程化实践」研究主线,反映出当前技术演进与商业落地的最新风向。涉及核心实体与技术议题包括:CoST、多标签学习、图像标注、软目标,读者可通过下方标签进入对应专题页查阅历史脉络与关联对比。

信源核验说明:本文原始线索采编自「arXiv cs.LG」(@Tong Jin, Meng-Long Wang, Xingyu Zhao, Sheng-Jun Huang),经 AITOP 多维评分管线结构化提炼。