{"title":"用视觉语言模型增强人再识别中的视觉分析。","authors":"Wang Xia, Tianci Wang, Jiawei Li, Guodao Sun, Haidong Gao, Xu Tan, Ronghua Liang","doi":"10.1109/MCG.2025.3593227","DOIUrl":null,"url":null,"abstract":"<p><p>Image-based person re-identification aims to match individuals across multiple cameras. Despite advances in machine learning, their effectiveness in real-world scenarios remains limited, often leaving users to handle fine-grained matching manually. Recent work has explored textual information as auxiliary cues, but existing methods generate coarse descriptions and fail to integrate them effectively into retrieval workflows. To address these issues, we adopt a vision-language model fine-tuned with domain-specific knowledge to generate detailed textual descriptions and keywords for pedestrian images. We then create a joint search space combining visual and textual information, using image clustering and keyword co-occurrence to build a semantic layout. Additionally, we introduce a dynamic spiral word cloud algorithm to improve visual presentation and enhance semantic associations. Finally, we conduct case studies, a user study, and expert feedback, demonstrating the usability and effectiveness of our system.</p>","PeriodicalId":55026,"journal":{"name":"IEEE Computer Graphics and Applications","volume":"PP ","pages":""},"PeriodicalIF":1.4000,"publicationDate":"2025-07-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Enhancing Visual Analysis in Person Re-Identification With Vision-Language Models.\",\"authors\":\"Wang Xia, Tianci Wang, Jiawei Li, Guodao Sun, Haidong Gao, Xu Tan, Ronghua Liang\",\"doi\":\"10.1109/MCG.2025.3593227\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p><p>Image-based person re-identification aims to match individuals across multiple cameras. Despite advances in machine learning, their effectiveness in real-world scenarios remains limited, often leaving users to handle fine-grained matching manually. Recent work has explored textual information as auxiliary cues, but existing methods generate coarse descriptions and fail to integrate them effectively into retrieval workflows. To address these issues, we adopt a vision-language model fine-tuned with domain-specific knowledge to generate detailed textual descriptions and keywords for pedestrian images. We then create a joint search space combining visual and textual information, using image clustering and keyword co-occurrence to build a semantic layout. Additionally, we introduce a dynamic spiral word cloud algorithm to improve visual presentation and enhance semantic associations. Finally, we conduct case studies, a user study, and expert feedback, demonstrating the usability and effectiveness of our system.</p>\",\"PeriodicalId\":55026,\"journal\":{\"name\":\"IEEE Computer Graphics and Applications\",\"volume\":\"PP \",\"pages\":\"\"},\"PeriodicalIF\":1.4000,\"publicationDate\":\"2025-07-28\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"IEEE Computer Graphics and Applications\",\"FirstCategoryId\":\"94\",\"ListUrlMain\":\"https://doi.org/10.1109/MCG.2025.3593227\",\"RegionNum\":4,\"RegionCategory\":\"计算机科学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q3\",\"JCRName\":\"COMPUTER SCIENCE, SOFTWARE ENGINEERING\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"IEEE Computer Graphics and Applications","FirstCategoryId":"94","ListUrlMain":"https://doi.org/10.1109/MCG.2025.3593227","RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"COMPUTER SCIENCE, SOFTWARE ENGINEERING","Score":null,"Total":0}
Enhancing Visual Analysis in Person Re-Identification With Vision-Language Models.
Image-based person re-identification aims to match individuals across multiple cameras. Despite advances in machine learning, their effectiveness in real-world scenarios remains limited, often leaving users to handle fine-grained matching manually. Recent work has explored textual information as auxiliary cues, but existing methods generate coarse descriptions and fail to integrate them effectively into retrieval workflows. To address these issues, we adopt a vision-language model fine-tuned with domain-specific knowledge to generate detailed textual descriptions and keywords for pedestrian images. We then create a joint search space combining visual and textual information, using image clustering and keyword co-occurrence to build a semantic layout. Additionally, we introduce a dynamic spiral word cloud algorithm to improve visual presentation and enhance semantic associations. Finally, we conduct case studies, a user study, and expert feedback, demonstrating the usability and effectiveness of our system.
期刊介绍:
IEEE Computer Graphics and Applications (CG&A) bridges the theory and practice of computer graphics, visualization, virtual and augmented reality, and HCI. From specific algorithms to full system implementations, CG&A offers a unique combination of peer-reviewed feature articles and informal departments. Theme issues guest edited by leading researchers in their fields track the latest developments and trends in computer-generated graphical content, while tutorials and surveys provide a broad overview of interesting and timely topics. Regular departments further explore the core areas of graphics as well as extend into topics such as usability, education, history, and opinion. Each issue, the story of our cover focuses on creative applications of the technology by an artist or designer. Published six times a year, CG&A is indispensable reading for people working at the leading edge of computer-generated graphics technology and its applications in everything from business to the arts.