Scholarlink

<p>This study investigated the localization accuracy of Grad-CAM as an Explainable AI (XAI) technique for industrial inspection, specifically evaluating it on the WM811K semiconductor wafer defect dataset — a large-scale, publicly available collection of 25,519 labeled wafer maps. Localization performance was measured using the Top-10% Intersection over Union (IoU) metric on a held-out test sample (<InlineEquation ID="IEq1"><EquationSource Format="TEX">\(N = 3828\)</EquationSource></InlineEquation>), yielding a baseline accuracy of 0.129. A consistent radial attribution bias was observed across both Grad-CAM and XGrad-CAM, with both methods producing structurally similar bias profiles, suggesting the bias reflects a property of the model’s learned representations rather than a gradient-weighting artifact. To address the radial attribution bias, the authors implemented a radial suppression method, which improved the average Top-10% IoU by 19.4% (from 0.129 to 0.155), verified through a two-tailed paired <i>t</i>-test (<InlineEquation ID="IEq2"><EquationSource Format="TEX">\(p \ll 0.001\)</EquationSource></InlineEquation>). Class-level analysis revealed significant improvements for edge-ring defects while other defect types remained stable. Finally, an oracle-based refinement demonstrated that geometric representation characteristics account for a substantial portion of residual localization errors, establishing an upper bound of 0.278 IoU.</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

Geometry-aware localization evaluation of grad-CAM for wafer map defect classification

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月24日

查看全文
  • Tushar Dudeja,
  • Prachi Sharma

摘要

This study investigated the localization accuracy of Grad-CAM as an Explainable AI (XAI) technique for industrial inspection, specifically evaluating it on the WM811K semiconductor wafer defect dataset — a large-scale, publicly available collection of 25,519 labeled wafer maps. Localization performance was measured using the Top-10% Intersection over Union (IoU) metric on a held-out test sample (\(N = 3828\)), yielding a baseline accuracy of 0.129. A consistent radial attribution bias was observed across both Grad-CAM and XGrad-CAM, with both methods producing structurally similar bias profiles, suggesting the bias reflects a property of the model’s learned representations rather than a gradient-weighting artifact. To address the radial attribution bias, the authors implemented a radial suppression method, which improved the average Top-10% IoU by 19.4% (from 0.129 to 0.155), verified through a two-tailed paired t-test (\(p \ll 0.001\)). Class-level analysis revealed significant improvements for edge-ring defects while other defect types remained stable. Finally, an oracle-based refinement demonstrated that geometric representation characteristics account for a substantial portion of residual localization errors, establishing an upper bound of 0.278 IoU.