Scholarlink

<p>We address the problem of cross-modal retrieval in presence of multi-view and multi-label data. For this, we present Multi-view Multi-label Canonical Correlation Analysis (or MVMLCCA), which is a generalization of Canonical Correlation Analysis (CCA) for multi-view data. While CCA relies on explicit pairings/associations of samples between two views (or modalities), MVMLCCA uses high-level semantic information (such as multi-label annotations) to establish correspondence across multiple (two or more) views without the need of an explicit pairing/association of samples across multiple views. We also present Fast MVMLCCA (FMVMLCCA), a computationally efficient version of MVMLCCA which is experimentally shown to be several times faster than MVMLCCA. Extensive experimental analyses on three popular multi-modal datasets (IAPRTC-12, MS-COCO and MIR-Flickr) demonstrate that the proposed approach offers much more flexibility than the related approaches without compromising on scalability and cross-modal retrieval performance.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Multi-view multi-label canonical correlation analysis for cross-modal multimedia retrieval

  • 期刊论文

  • 发表日期：

    2025年7月31日

查看全文
  • Asha Rani,
  • Rushil Sanghavi,
  • Yashaswi Verma

摘要

We address the problem of cross-modal retrieval in presence of multi-view and multi-label data. For this, we present Multi-view Multi-label Canonical Correlation Analysis (or MVMLCCA), which is a generalization of Canonical Correlation Analysis (CCA) for multi-view data. While CCA relies on explicit pairings/associations of samples between two views (or modalities), MVMLCCA uses high-level semantic information (such as multi-label annotations) to establish correspondence across multiple (two or more) views without the need of an explicit pairing/association of samples across multiple views. We also present Fast MVMLCCA (FMVMLCCA), a computationally efficient version of MVMLCCA which is experimentally shown to be several times faster than MVMLCCA. Extensive experimental analyses on three popular multi-modal datasets (IAPRTC-12, MS-COCO and MIR-Flickr) demonstrate that the proposed approach offers much more flexibility than the related approaches without compromising on scalability and cross-modal retrieval performance.