计算机应用研究2024，Vol.41Issue(6)：1610-1617,8.DOI:10.19734/j.issn.1001-3695.2023.09.0420
面向分布式复杂数据样本的联邦语义分割方法综述
Survey on federated semantic segmentation methods for distributed complex data samples
摘要
Abstract
Semantic segmentation plays a crucial role in various fields such as medical image analysis and battlefield situatio-nal awareness.However,a single client often cannot provide a sufficient quantity and diversity of training data for the model.Therefore,it is necessary to train semantic segmentation models from distributed data,which exhibits complex and diverse characteristics.To prevent data privacy breaches and safeguard data security,the application of federated learning in the col-laborative training of semantic segmentation models across multiple clients has become a hot research topic in the field.Buil-ding upon the definition of federated semantic segmentation,this paper conducted a comprehensive analysis around the key characteristics of data heterogeneity and label deficiency in distributed complex data samples.The study encompassed a review of issues,methods,and exemplary model instances in federated semantic segmentation,evaluating the applicability and cha-racteristics of different methods,summarizing current application outcomes.The paper also proposed potential research oppor-tunities to address the issues of data heterogeneity and label deficiency.The research provides insights and references for the development of federated semantic segmentation methods and related studies tailored for distributed complex data samples.关键词
语义分割/联邦学习/协同训练/数据异质性/标签缺失Key words
semantic segmentation/federated learning/collaborative training/data heterogeneity/label deficiency分类
信息技术与安全科学引用本文复制引用
董成荣,姚俊萍,李晓军,苏逸,周志杰..面向分布式复杂数据样本的联邦语义分割方法综述[J].计算机应用研究,2024,41(6):1610-1617,8.基金项目
国家自然科学基金资助项目(61833016,62227814) （61833016,62227814）
陕西省科技创新团队项目(2022TD-24) （2022TD-24）
