Scholarlink

<p>Existing fabric datasets consist of isolated and non-time-series images, failing to reflect the complexities and variations present in real-world industrial environments. To support fabric defect detection in real-world industrial environments, we present a high-quality <b>T</b>ime-<b>S</b>eries fabric dataset (<b>TSfabrics</b>). By capturing time-series fabric images, <b>TSfabrics</b> reflects the continuous and complex nature of actual fabric manufacturing <b>on circular knitting machines</b>. <b>TSfabrics</b> encompasses 22 continuous fabric production scenarios and consists of 93196 grayscale fabric images collected under diverse production conditions, including different illuminations, production speeds, and fabric types. In addition, <b>TSfabrics</b> contains defect-free and defective samples, with pixel-level annotations that differentiate true defects from production-related “cutlines” found in defect-free fabric images. Meanwhile, <b>TSfabrics</b> covers a wide range of fabric defect types. To sum up, <b>TSfabrics</b> offers a valuable resource for advancing practical fabric defect detection methods and bridging the gap between academic research and industrial application. We believe that <b>TSfabrics</b> enables the development and evaluation of defect detection models that are better suited for deployment on continuous production lines.</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

TSFabrics: A Time-Series Fabric Dataset for Real-Time Defect Detection on Circular Knitting Machines

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月7日

查看全文
  • Yan-Qin Ni,
  • Pei-Kai Huang,
  • Wei-Jen Wang,
  • Deron Liang,
  • Chia-Yu Lin

摘要

Existing fabric datasets consist of isolated and non-time-series images, failing to reflect the complexities and variations present in real-world industrial environments. To support fabric defect detection in real-world industrial environments, we present a high-quality Time-Series fabric dataset (TSfabrics). By capturing time-series fabric images, TSfabrics reflects the continuous and complex nature of actual fabric manufacturing on circular knitting machines. TSfabrics encompasses 22 continuous fabric production scenarios and consists of 93196 grayscale fabric images collected under diverse production conditions, including different illuminations, production speeds, and fabric types. In addition, TSfabrics contains defect-free and defective samples, with pixel-level annotations that differentiate true defects from production-related “cutlines” found in defect-free fabric images. Meanwhile, TSfabrics covers a wide range of fabric defect types. To sum up, TSfabrics offers a valuable resource for advancing practical fabric defect detection methods and bridging the gap between academic research and industrial application. We believe that TSfabrics enables the development and evaluation of defect detection models that are better suited for deployment on continuous production lines.