Scholarlink

<p>Underwater object detection faces challenges due to poor underwater image quality, significant multi-scale variations in targets, and severe occlusion. To tackle these issues, we propose a novel model integrating multi-scale attention and adaptive feature fusion to enhance detection accuracy. Leveraging the DEIM framework, we propose a Multi-Scale Attention Block (MSAB) to improve feature extraction across scales, a Lightweight Sparse Self-Attention Block (LSSA) for noise suppression, an Adaptive Weighted Downsampling Block (AWDS) to preserve information, and a Context-Guided Feature Fusion Module (CGFM) for intelligent feature integration. Experimental results on the URPC, DUO, and RUOD datasets demonstrate significant improvements in AP, AP50, and AP75 metrics compared to the baseline model DEIM, with gains of 3.2%, 3.3%, and 3.4% on URPC, 3.9%, 3.6%, and 4.7% on DUO, and 2.3%, 2.2%, and 2.6% on RUOD, respectively. Moreover, extensive visualization experiments consistently demonstrate that our model robustly handles the inherent complexities of underwater environments. The code and datasets are publicly available at <a href="https://github.com/EUOD/MSAuod">https://github.com/EUOD/MSAuod</a>, accompanied by detailed usage guides to facilitate reproducibility.</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

Enhanced underwater object detection via multi-scale attention and adaptive feature fusion

  • 期刊论文

  • 发表日期：

    2026年7月27日

查看全文
  • Hui Chen,
  • Yongjie Yu,
  • Tao Fu

摘要

Underwater object detection faces challenges due to poor underwater image quality, significant multi-scale variations in targets, and severe occlusion. To tackle these issues, we propose a novel model integrating multi-scale attention and adaptive feature fusion to enhance detection accuracy. Leveraging the DEIM framework, we propose a Multi-Scale Attention Block (MSAB) to improve feature extraction across scales, a Lightweight Sparse Self-Attention Block (LSSA) for noise suppression, an Adaptive Weighted Downsampling Block (AWDS) to preserve information, and a Context-Guided Feature Fusion Module (CGFM) for intelligent feature integration. Experimental results on the URPC, DUO, and RUOD datasets demonstrate significant improvements in AP, AP50, and AP75 metrics compared to the baseline model DEIM, with gains of 3.2%, 3.3%, and 3.4% on URPC, 3.9%, 3.6%, and 4.7% on DUO, and 2.3%, 2.2%, and 2.6% on RUOD, respectively. Moreover, extensive visualization experiments consistently demonstrate that our model robustly handles the inherent complexities of underwater environments. The code and datasets are publicly available at https://github.com/EUOD/MSAuod, accompanied by detailed usage guides to facilitate reproducibility.