Scholarlink

<p>With the rapid advancement of deep learning, integrating segmentation techniques into scene text detection has gained recognition, yet industrial chip text detection faces tough challenges–dense distribution, metallic reflections, low-resolution characters–that existing methods struggle to solve. In this study, we propose GeneRetinaNet, a specialized chip text localization framework with three core modules: Graph-based Recursive Convolution (GRConv) enhances multi-scale feature capture to mitigate fragmentation; Local Transformation CopyPaste (LTCP) generates chip-specific augmented samples; Weighted Alignment Pyramid Network (WAFPN) reduces cross-scale semantic gaps via dynamic weighting. Our method performs well on ChipCode (a self-calibrated industrial dataset) and is competitive on ICDAR2015, offering a targeted solution for industrial chip text localization.</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

GeneRetinaNet: a weighted aligned pyramid structure chip text localization network based on Graph-based Recursive Convolution

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年11月4日

查看全文
  • Jie Cao,
  • Song Cai,
  • Daolong Han,
  • Chao Dong,
  • Jianfeng Xu,
  • Guihua Lu

摘要

With the rapid advancement of deep learning, integrating segmentation techniques into scene text detection has gained recognition, yet industrial chip text detection faces tough challenges–dense distribution, metallic reflections, low-resolution characters–that existing methods struggle to solve. In this study, we propose GeneRetinaNet, a specialized chip text localization framework with three core modules: Graph-based Recursive Convolution (GRConv) enhances multi-scale feature capture to mitigate fragmentation; Local Transformation CopyPaste (LTCP) generates chip-specific augmented samples; Weighted Alignment Pyramid Network (WAFPN) reduces cross-scale semantic gaps via dynamic weighting. Our method performs well on ChipCode (a self-calibrated industrial dataset) and is competitive on ICDAR2015, offering a targeted solution for industrial chip text localization.