A Pix2pixGAN-Based Method For Carbide Segmentation In GCr15 Steel.

Saved in:

Bibliographic Details
Title:	A Pix2pixGAN-Based Method For Carbide Segmentation In GCr15 Steel.
Authors:	Wang, Jiangang¹ wm094212@163.com, Lian, Xiaolong² 13102731601@163.com, Han, Borui² 1491453212@qq.com, Sun, Yixiao³ robinsunyixiao@163.com, Ju, Dongying¹ dyju.sitec@gmail.com, Wu, Yanzhao⁴ yzw@hebust.edu.cn, Zhang, Xin⁵ zhxin5210@163.com, Yang, Liyong⁶ sgyangliyong@hbisco.com
Source:	IAENG International Journal of Computer Science. Jun2026, Vol. 53 Issue 6, p2264-2271. 8p.
Subjects:	Generative adversarial networks, Carbides, Artificial neural networks, Metallography, Deep learning, Data augmentation, Steel
Abstract:	Metallographic analysis plays a key role in modern materials science, as the morphology and distribution of carbides after quenching strongly influence the performance and its subsequent heat treatment. In this study, a dataset of quenched alloy carbides was established, and a carbide segmentation model based on pix2pix Generative Adversarial Network (pix2pixGAN) was proposed. The model integrates a new feature enhancement module, ECA-NLA, which combines Efficient Channel Attention mechanism with non-local spatial attention module to strengthen feature extraction, enhance channel perception, and enable adaptive feature weighting. In addition, the Inception v1 module was incorporated to enable multi-scale feature extraction and reduce pixel-level information loss, while depthwise separable convolutions were used to improve the generator's representational capacity and efficiency, resulting in notable performance gains. Experimental results demonstrate the effectiveness of the improved pix2pixGAN model, achieving an Intersection over Union of 83.8% on the proposed dataset, which is 5.1% higher than that of the baseline pix2pixGAN model. By leveraging recent advances in deep learning architectures and data optimization techniques, this study improves the automation and accuracy of alloy carbide evaluation and provides a promising solution for automated metallographic analysis. [ABSTRACT FROM AUTHOR]
	Copyright of IAENG International Journal of Computer Science is the property of International Association of Engineers (IAENG) and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database:	Engineering Source

Description
Abstract:	Metallographic analysis plays a key role in modern materials science, as the morphology and distribution of carbides after quenching strongly influence the performance and its subsequent heat treatment. In this study, a dataset of quenched alloy carbides was established, and a carbide segmentation model based on pix2pix Generative Adversarial Network (pix2pixGAN) was proposed. The model integrates a new feature enhancement module, ECA-NLA, which combines Efficient Channel Attention mechanism with non-local spatial attention module to strengthen feature extraction, enhance channel perception, and enable adaptive feature weighting. In addition, the Inception v1 module was incorporated to enable multi-scale feature extraction and reduce pixel-level information loss, while depthwise separable convolutions were used to improve the generator's representational capacity and efficiency, resulting in notable performance gains. Experimental results demonstrate the effectiveness of the improved pix2pixGAN model, achieving an Intersection over Union of 83.8% on the proposed dataset, which is 5.1% higher than that of the baseline pix2pixGAN model. By leveraging recent advances in deep learning architectures and data optimization techniques, this study improves the automation and accuracy of alloy carbide evaluation and provides a promising solution for automated metallographic analysis. [ABSTRACT FROM AUTHOR]
ISSN:	1819656X