Residual neural network-based fully convolutional network for microstructure segmentation

Jang, Junmyoung; Van, Donghyun; Jang, Hyojin; Baik, Dae Hyun; Yoo, Sang Duk; Park, Jaewoong; Mhin, Sungwook; Mazumder, Jyoti; Lee, S. H.

doi:10.1080/13621718.2019.1687635

Cited by 26 publications

(9 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Azimi et al reported the successful implementation of VGGNet 17 , which is a pretrained CNN proposed by Krizhevsky, for classifying microstructures from LOM and SEM images of a steel 18 . Since then, a number of studies applying CNN have been conducted such as the application of DenseNet 19 to detect defects in steels 20 and ResNet18 21 to classify microstructures of welded steels 22 . It was verified that the performance of CNN-based methods is as good as that of humans.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Unsupervised microstructure segmentation by mimicking metallurgists’ approach to pattern recognition

Kim

Inoue

Kasuya

2020

Sci Rep

View full text Add to dashboard Cite

An efficient deep learning method is presented for distinguishing microstructures of a low carbon steel. There have been numerous endeavors to reproduce the human capability of perceptually classifying different textures using machine learning methods, but this is still very challenging owing to the need for a vast labeled image dataset. In this study, we introduce an unsupervised machine learning technique based on convolutional neural networks and a superpixel algorithm for the segmentation of a low-carbon steel microstructure without the need for labeled images. The effectiveness of the method is demonstrated with optical microscopy images of steel microstructures having different patterns taken at different resolutions. In addition, several evaluation criteria for unsupervised segmentation results are investigated along with the hyperparameter optimization.

show abstract

Section: Introductionmentioning

confidence: 99%

“…2 Graduate School of Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo, Tokyo 113-8656, Japan. * email: inoue@ material.t.u-tokyo.ac.jp www.nature.com/scientificreports/ to detect defects in steels 20 and ResNet18 21 to classify microstructures of welded steels 22 . It was verified that the performance of CNN-based methods is as good as that of humans.…”

mentioning

confidence: 99%

Unsupervised microstructure segmentation by mimicking metallurgists’ approach to pattern recognition

Kim

Inoue

Kasuya

2020

Sci Rep

View full text Add to dashboard Cite

show abstract

“…Roberts et al 18 proposed a DenseNet based U-Net model for semantic segmentation of metallographic images. Jang et al 19 proposed a ResNet-based U-Net segmentation network. Both of these approaches are well-known in the literature on semantic segmentation 12 .…”

Section: Related Workmentioning

confidence: 99%

Microstructural segmentation using a union of attention guided U-Net models with different color transformed images

Biswas

Pramanik

Sen

et al. 2023

Sci Rep

View full text Add to dashboard Cite

Metallographic images or often called the microstructures contain important information about metals, such as strength, toughness, ductility, corrosion resistance, which are used to choose the proper materials for various engineering applications. Thus by understanding the microstructures, one can determine the behaviour of a component made of a particular metal, and can predict the failure of that component in certain conditions. Image segmentation is a powerful technique for determination of morphological features of the microstructure like volume fraction, inclusion morphology, void, and crystal orientations. These are some key factors for determining the physical properties of metal. Therefore, automatic micro-structure characterization using image processing is useful for industrial applications which currently adopts deep learning-based segmentation models. In this paper, we propose a metallographic image segmentation method using an ensemble of modified U-Nets. Three U-Net models having the same architecture are separately fed with color transformed imaged (RGB, HSV and YUV). We improvise the U-Net with dilated convolutions and attention mechanisms to get finer grained features. Then we apply the sum-rule-based ensemble method on the outcomes of U-Net models to get the final prediction mask. We achieve the mean intersection over union (IoU) score of 0.677 on a publicly available standard dataset, namely MetalDAM. We also show that the proposed method obtains results comparable to state-of-the-art methods with fewer number of model parameters. The source code of the proposed work can be found at https://github.com/mb16biswas/attention-unet.

show abstract

“…For example, a type of CNN was used for segmenting the microstructures of welded carbon steel to measure the fraction of acicular ferrite. 108 A segmentation algorithm derived from U-Net 90 and DenseNet 94 was used for semantic segmentation of three common crystallographic defects in structural alloys. 109 Image segmentation is typically supervised learning where the training data are created manually via labor-intensive methods.…”

Section: Semantic Segmentationmentioning

confidence: 99%

A Review of Application of Machine Learning in Design, Synthesis, and Characterization of Metal Matrix Composites: Current Status and Emerging Applications

Kordijazi

Zhao

Zhang

et al. 2021

JOM

View full text Add to dashboard Cite

In this article we provide an overview on the current and emerging applications of machine learning (ML) in the design, synthesis, and characterization of metal matrix composites (MMC). We have demonstrated that ML methods can be applied in three distinct categories, namely property prediction, microstructure analysis, and process optimization, which are associated with three major classes of ML techniques, i.e., regression, classification, and optimal control, respectively. ML algorithms have been successfully applied for prediction of mechanical, tribological, corrosion, and wetting properties of different MMCs. However, ML methods (e.g., computer vision, which is suitable for microstructural characterization and defect detections) and optimization algorithms (e.g., reinforcement learning) have not been fully utilized for design, processing, and characterization of metal matrix composites despite their enormous capacities. We conclude that ML methods are promising not only to predict various properties but also to automate microstructural analysis and optimization of manufacturing MMCs.

show abstract

Residual neural network-based fully convolutional network for microstructure segmentation

Cited by 26 publications

References 27 publications

Unsupervised microstructure segmentation by mimicking metallurgists’ approach to pattern recognition

Unsupervised microstructure segmentation by mimicking metallurgists’ approach to pattern recognition

Microstructural segmentation using a union of attention guided U-Net models with different color transformed images

A Review of Application of Machine Learning in Design, Synthesis, and Characterization of Metal Matrix Composites: Current Status and Emerging Applications

Contact Info

Product

Resources

About