Scholarlink

<p>Multi-branch methods for person re-identification mainly extract features through separate branches, but often ignore correlations between them. We propose a Multi-Branch Attention Feature Fusion Network (MAFFN) to address this limitation. MAFFN contains three branches: attention, global, and local. The attention branch uses a Parallel Attention Module that combines channel and spatial attention. The global branch captures overall features, while the local branch learns fine-grained details. A feature fusion strategy strengthens correlations between branches, improving feature discrimination and robustness. Experimental evaluations conducted on three publicly available datasets validate the superiority of the MAFFN over other state-of-the-art methods. The code is available at <a href="https://github.com/JL0424/MAFFN.git">https://github.com/JL0424/MAFFN.git</a>.</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-branch attention feature fusion network for person re-identification

  • 期刊论文

  • 发表日期：

    2025年10月16日

查看全文
  • Ke Han,
  • Long Jin,
  • Junpeng Yang,
  • Zongwang Lv

摘要

Multi-branch methods for person re-identification mainly extract features through separate branches, but often ignore correlations between them. We propose a Multi-Branch Attention Feature Fusion Network (MAFFN) to address this limitation. MAFFN contains three branches: attention, global, and local. The attention branch uses a Parallel Attention Module that combines channel and spatial attention. The global branch captures overall features, while the local branch learns fine-grained details. A feature fusion strategy strengthens correlations between branches, improving feature discrimination and robustness. Experimental evaluations conducted on three publicly available datasets validate the superiority of the MAFFN over other state-of-the-art methods. The code is available at https://github.com/JL0424/MAFFN.git.