{"title":"Intrinsic feature extraction for unsupervised domain adaptation","authors":"Xinzhi Cao, Yinsai Guo, Wenbin Yang, Xiangfeng Luo, Shaorong Xie","doi":"10.1108/ijwis-04-2023-0062","DOIUrl":null,"url":null,"abstract":"\nPurpose\nUnsupervised domain adaptation object detection not only mitigates model terrible performance resulting from domain gap, but also has the ability to apply knowledge trained on a definite domain to a distinct domain. However, aligning the whole feature may confuse the object and background information, making it challenging to extract discriminative features. This paper aims to propose an improved approach which is called intrinsic feature extraction domain adaptation (IFEDA) to extract discriminative features effectively.\n\n\nDesign/methodology/approach\nIFEDA consists of the intrinsic feature extraction (IFE) module and object consistency constraint (OCC). The IFE module, designed on the instance level, mainly solves the issue of the difficult extraction of discriminative object features. Specifically, the discriminative region of the objects can be paid more attention to. Meanwhile, the OCC is deployed to determine whether category prediction in the target domain brings into correspondence with it in the source domain.\n\n\nFindings\nExperimental results demonstrate the validity of our approach and achieve good outcomes on challenging data sets.\n\n\nResearch limitations/implications\nLimitations to this research are that only one target domain is applied, and it may change the ability of model generalization when the problem of insufficient data sets or unseen domain appeared.\n\n\nOriginality/value\nThis paper solves the issue of critical information defects by tackling the difficulty of extracting discriminative features. And the categories in both domains are compelled to be consistent for better object detection.\n","PeriodicalId":44153,"journal":{"name":"International Journal of Web Information Systems","volume":null,"pages":null},"PeriodicalIF":2.5000,"publicationDate":"2023-07-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"2","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"International Journal of Web Information Systems","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1108/ijwis-04-2023-0062","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"COMPUTER SCIENCE, INFORMATION SYSTEMS","Score":null,"Total":0}
引用次数: 2
Abstract
Purpose
Unsupervised domain adaptation object detection not only mitigates model terrible performance resulting from domain gap, but also has the ability to apply knowledge trained on a definite domain to a distinct domain. However, aligning the whole feature may confuse the object and background information, making it challenging to extract discriminative features. This paper aims to propose an improved approach which is called intrinsic feature extraction domain adaptation (IFEDA) to extract discriminative features effectively.
Design/methodology/approach
IFEDA consists of the intrinsic feature extraction (IFE) module and object consistency constraint (OCC). The IFE module, designed on the instance level, mainly solves the issue of the difficult extraction of discriminative object features. Specifically, the discriminative region of the objects can be paid more attention to. Meanwhile, the OCC is deployed to determine whether category prediction in the target domain brings into correspondence with it in the source domain.
Findings
Experimental results demonstrate the validity of our approach and achieve good outcomes on challenging data sets.
Research limitations/implications
Limitations to this research are that only one target domain is applied, and it may change the ability of model generalization when the problem of insufficient data sets or unseen domain appeared.
Originality/value
This paper solves the issue of critical information defects by tackling the difficulty of extracting discriminative features. And the categories in both domains are compelled to be consistent for better object detection.
期刊介绍:
The Global Information Infrastructure is a daily reality. In spite of the many applications in all domains of our societies: e-business, e-commerce, e-learning, e-science, and e-government, for instance, and in spite of the tremendous advances by engineers and scientists, the seamless development of Web information systems and services remains a major challenge. The journal examines how current shared vision for the future is one of semantically-rich information and service oriented architecture for global information systems. This vision is at the convergence of progress in technologies such as XML, Web services, RDF, OWL, of multimedia, multimodal, and multilingual information retrieval, and of distributed, mobile and ubiquitous computing. Topicality While the International Journal of Web Information Systems covers a broad range of topics, the journal welcomes papers that provide a perspective on all aspects of Web information systems: Web semantics and Web dynamics, Web mining and searching, Web databases and Web data integration, Web-based commerce and e-business, Web collaboration and distributed computing, Internet computing and networks, performance of Web applications, and Web multimedia services and Web-based education.