"Enet X-rays image"

Read full text
Authors: Zihao Gong
Cited by: 0

Abstract

"Relying on the actual production scenario of a steel pipe manufacturing enterprise, this paper constructs a real-world weld DR image dataset named Degraded Steel Pipe (DSP). This dataset consists of 330 real industrial DR images with varying sizes, fully retaining typical weld degradation features, including but not limited to, incomplete penetration, cracks, and porosity. To provide a more challenging evaluation environment, the dataset is augmented with various auxiliary markers, such as alphanumeric characters. This allows for a robust assessment of enhancement algorithms, particularly in scenarios involving high noise levels, weak textures, and mixed content. The public release of this dataset is the first of its kind, addressing the critical lack of real-world, unpaired data for industrial weld X-ray image enhancement. It establishes a standardized benchmark for subsequent research and empirically validates the efficacy and practical utility of our proposed method under realistic industrial conditions."

Fields of Study & Subjects

Image Enhancement Techniques Industrial Vision Systems and Defect Detection Non-Destructive Testing Techniques

Publication details

Journal / source
IEEE DataPort
Publication date
1 March 2026
Language
EN
Document type
Dataset