The IRMA Community
Newsletters
Research IRM
Click a keyword to search titles using our InfoSci-OnDemand powered search:
|
AI-Powered Defect Detection Using Deep Learning Enhancing Industrial Automation with U-Net and YOLOv3
Abstract
This chapter explores the application of deep learning models, specifically U-Net and YOLOv3, in detecting crack-type defects in industrial materials. The research addresses key challenges, such as the limited availability of defective data samples, the difficulty of distinguishing fine defect details, and the need for precise quantitative evaluations in automated inspection systems. The chapter introduces the significance of defect detection in the context of smart manufacturing and discusses the limitations of traditional manual inspection methods. U-Net, known for its superior pixel-level semantic segmentation capabilities, is utilized for detailed defect localization, while YOLOv3, a real-time object detection model, is optimized for fast detection but has trade-offs in precision, particularly for irregular defect shapes. Both models are trained on a custom dataset provided by a material manufacturer incorporating a novel augmentation method that enhances generalization and improves detection accuracy by expanding defective sample diversity.
Related Content
|
Mohammed Adi Al Battashi, Mohamad A. M. Adnan, Asyraf Isyraqi Bin Jamil, Majid Adi Al-Battashi.
© 2026.
30 pages.
|
|
Potchong M. Jackaria, Al-adzran G. Sali, Hana An L. Alvarado, Rashidin H. Moh. Jiripa, Al-sabrie Y. Sahijuan.
© 2026.
26 pages.
|
|
Elizabeth Gross.
© 2026.
30 pages.
|
|
Siti Nazleen Abdul Rabu, Xie Fengli, Ng Man Yi.
© 2026.
44 pages.
|
|
Mohammed Abdul Wajeed.
© 2026.
30 pages.
|
|
Aldammien A. Sukarno, Al-adzkhan N. Abdulbarie, Wati Sheena M. Bulkia, Potchong M. Jackaria.
© 2026.
24 pages.
|
|
Abdulla Sultan Binhareb Almheiri, Humaid Albastaki, Hanadi Alrashdan.
© 2026.
26 pages.
|
|
|