自动化学报2025，Vol.51Issue(6)：1191-1204,14.DOI:10.16383/j.aas.c240636
基于图像特定分类器的弱监督语义分割
Image-specific Classifiers for Weakly Supervised Semantic Segmentation
摘要
Abstract
Weakly supervised semantic segmentation algorithms based on image-level labels have garnered wide-spread attention due to their low annotation costs.These algorithms utilize class activation maps(CAMs)gener-ated by classification networks to convert from image-level labels to pixel-level labels.However,CAMs only focus on the most discriminative regions in the image,resulting in a large gap between pseudo-labels generated by CAMs and ground truth.The gap includes under-activation and mis-activation issues.Under-activation arises from excessive intra-class differences in the dataset,making a single classifier insufficient to accurately identify all pixels of the same category,while mis-activation occurs when inter-class differences are too small,preventing the classifier from effectively distinguishing pixels of different categories.This paper considers that the intra-class difference for pixels of the same class within an image is smaller than that of the dataset,so we design the image-specific classifier based on the class center to enhance the recognition capability for pixels of the same class,thereby addressing the under-activation issue.Besides,considering that the class center serves as the representative of class in the feature space,we design the class center constrained loss function.By enlarging the distances of different class centers,this func-tion indirectly pushes apart the feature distributions of different classes,thereby mitigating the issue of mis-activa-tion.The proposed image-specific classifier can be plugged into other weakly supervised semantic segmentation net-works,which can replace the classifiers of classification networks to generate higher quality CAMs.Experimental results demonstrate the superior performance of the proposed methods on two benchmark datasets,validating the effectiveness of this method.关键词
语义分割/图像级标签/分类器/类激活图/弱监督学习Key words
Semantic segmentation/image-level label/classifiers/class activation maps/weakly supervised learning引用本文复制引用
郭子麟,吴东岳,高常鑫,桑农..基于图像特定分类器的弱监督语义分割[J].自动化学报,2025,51(6):1191-1204,14.基金项目
国家自然科学基金(62176097,61433007),中央高校基本科研业务费(2019kfyXKJC024),计算智能与智能控制111计划(B18024)资助Supported by National Natural Science Foundation of China(62176097,61433007),Fundamental Research Funds for the Central Universities(2019kfyXKJC024),and 111 Project on Computational Intelligence and Intelligent Control(B18024) （62176097,61433007）
