Self-Net: A Dynamically Configurable U-Net Variant for Real-Time Steel Surface Defect Detection

Advances in transdisciplinary engineering 2026 Book chapter Open access

Read full text
Authors: Xiaofang Shi, Shuoxuan Wang, Zhihua Xiang, Jiajia Gao, Jingyi Liu, Ting Huang, Xinyi Guo, Junhao Wang
DOI: 10.3233/atde251698
Cited by: 0

Abstract

To address the dual requirements of real-time performance and multi-class recognition accuracy in industrial steel surface defect detection, this study proposes Self-Net, a dynamically configurable segmentation model. The task is challenged by complex background interference (e.g., oil stains, oxide scales), tiny defects (some

Fields of Study & Subjects

Industrial Vision Systems and Defect Detection Infrastructure Maintenance and Monitoring Non-Destructive Testing Techniques

Publication details

Journal / source
Advances in transdisciplinary engineering
Publication date
4 February 2026
Language
EN
Document type
Book chapter