A Study on the Fabric Defect Analysis Using Machine AI Deep Learning Models 


Vol. 62,  No. 6, pp. 402-410, Dec.  2025
10.12772/TSE.2025.62.402


PDF Full-Text
  Abstract

This study presents a comparative analysis of representative object detectionbased AI deep learning models(YOLOv4, YOLOX, RT-DETR) to enhance the accuracy and real-time performance of fabric defect detection, a critical step in textile manufacturing processes. Experimental results indicate that RT-DETR outperforms the other models in Accuracy, Recall, and F1-Score, with the exception of Precision. In particular, RT-DETR demonstrates a significantly higher Recall, suggesting its suitability for product categories where defect omission is unacceptable. This superiority appears to stem from its Transformer-based architecture, which effectively captures global features in textile images characterized by repetitive patterns and complex surface textures. However, RT-DETR exhibits a maximum processing speed of approximately 130 FPS, which is lower compared to the other models. While it is sufficient for general manufacturing processes, it may impose limitations in ultra-high-speed production environments. Overall, the findings confirm that the Transformer-based RT-DETR model represents the most suitable option for fabric defect analysis using AI deep learning techniques.

  Statistics
Cumulative Counts from November, 2022
Multiple requests among the same browser session are counted as one view. If you mouse over a chart, the values of data points will be shown.


  Cite this article

[IEEE Style]

최현석, 이한건, 이영훈, "A Study on the Fabric Defect Analysis Using Machine AI Deep Learning Models," Textile Science and Engineering, vol. 62, no. 6, pp. 402-410, 2025. DOI: 10.12772/TSE.2025.62.402.

[ACM Style]

최현석, 이한건, and 이영훈. 2025. A Study on the Fabric Defect Analysis Using Machine AI Deep Learning Models. Textile Science and Engineering, 62, 6, (2025), 402-410. DOI: 10.12772/TSE.2025.62.402.

Search
(IN TITLE, AUTHOR, ABSTRACT,KEYWORDS)

Advanced Search ▼

POPULAR KEYWORDS
(TOP 10 KEYWORDS)

Recent Publications
(LAST 3 YEARS)

pISSN : 1225 - 1089
eISSN : 2288 - 6419

한국연구재단 우수등재학술지 유지 (2022년)