电子学报2025，Vol.53Issue(6)：1741-1754,14.DOI:10.12263/DZXB.20250024
类感知对比学习的弱监督语义分割
Class-Aware Contrastive Learning for Weakly Supervised Semantic Segmentation
摘要
Abstract
In image-level weakly supervised semantic segmentation(WSSS),class activation map(CAM)are com-monly used to localize object regions.However,existing methods often encounter challenges such as under-activation in ob-ject regions and erroneous activation in background regions when generating CAM.This paper proposes a class-aware con-trastive learning(CA-CL)framework for weakly supervised semantic segmentation,which significantly enhances the mod-el's ability to accurately localize object regions by integrating text prompts and image category information.Firstly,we ana-lyze the influence of different text prompt templates on the class activation maps of various categories,on this basis,to ob-tain more adaptive class representations,we construct a contextual prompt set and design a dynamic contextual prompt se-lection strategy.This strategy generates the most appropriate contextual prompts based on the similarity between image ob-ject regions and text prompts.Secondly,we adopt an image-text contrastive learning approach to enhance the model's per-formance in aligning image and text semantics,and we design a contrastive loss function to guide the model training pro-cess.Finally,we introduce a class-specific background suppression module to mitigate erroneous activation in background regions closely related to object categories,thereby generating more complete and compact class activation maps and achieving more precise semantic segmentation.Experiments conducted on benchmark datasets PASCAL VOC 2012 and MS COCO 2014 demonstrate the effectiveness of the proposed framework,achieving mIoU values of 71.9％and 43.9％,re-spectively.The results demonstrate superior performance compared to existing methods,significantly improving the accura-cy of weakly supervised semantic segmentation.关键词
弱监督语义分割/类激活图/类感知/对比学习/文本提示Key words
weakly supervised semantic segmentation/class activation maps/class-aware/contrastive learning/text prompt分类
信息技术与安全科学引用本文复制引用
白雪飞,许文杰,王渊辉,王文剑..类感知对比学习的弱监督语义分割[J].电子学报,2025,53(6):1741-1754,14.基金项目
国家自然科学基金(No.U21A20513,No.62476157) （No.U21A20513,No.62476157）
太行山西省实验室技术攻关专项资助项目(No.THYF-JSZX-24010200) National Natural Science Foundation of China(No.U21A20513,No.62476157) （No.THYF-JSZX-24010200）
Key Technolo-gies Program of Taihang Laboratory in Shanxi Province(No.THYF-JSZX-24010200) （No.THYF-JSZX-24010200）
