基于模糊边界的深度神经网络图像分割损失函数

IF 0.5 4区工程技术 Q4 ENGINEERING, ELECTRICAL & ELECTRONIC

Electronics and Communications in Japan Pub Date : 2023-10-04 DOI:10.1002/ecj.12429

Yuma Hakumura, Taiyo Ito, Shiori Matsui, Yuya Akiba, Kimiya Aoki, Yuki Nakashima, Kiyoshi Hirao, Manabu Fukushima

{"title":"基于模糊边界的深度神经网络图像分割损失函数","authors":"Yuma Hakumura, Taiyo Ito, Shiori Matsui, Yuya Akiba, Kimiya Aoki, Yuki Nakashima, Kiyoshi Hirao, Manabu Fukushima","doi":"10.1002/ecj.12429","DOIUrl":null,"url":null,"abstract":"<p>This study deals with the task of segmentation of SEM images of fine ceramics sintered bodies by using deep neural network (DNN). In particular, we focus on misclassification caused by the blurriness of grain boundaries(boundaries between particles). Therefore, we utilize the frequency distribution of brightness gradient of grain boundaries and give higher weights to pixels with lower gradient values. Experiments confirmed that the model trained with proposed loss function gave the best prediction results.</p>","PeriodicalId":50539,"journal":{"name":"Electronics and Communications in Japan","volume":"106 4","pages":""},"PeriodicalIF":0.5000,"publicationDate":"2023-10-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Loss function for ambiguous boundaries for deep neural network (DNN) for image segmentation\",\"authors\":\"Yuma Hakumura, Taiyo Ito, Shiori Matsui, Yuya Akiba, Kimiya Aoki, Yuki Nakashima, Kiyoshi Hirao, Manabu Fukushima\",\"doi\":\"10.1002/ecj.12429\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p>This study deals with the task of segmentation of SEM images of fine ceramics sintered bodies by using deep neural network (DNN). In particular, we focus on misclassification caused by the blurriness of grain boundaries(boundaries between particles). Therefore, we utilize the frequency distribution of brightness gradient of grain boundaries and give higher weights to pixels with lower gradient values. Experiments confirmed that the model trained with proposed loss function gave the best prediction results.</p>\",\"PeriodicalId\":50539,\"journal\":{\"name\":\"Electronics and Communications in Japan\",\"volume\":\"106 4\",\"pages\":\"\"},\"PeriodicalIF\":0.5000,\"publicationDate\":\"2023-10-04\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Electronics and Communications in Japan\",\"FirstCategoryId\":\"5\",\"ListUrlMain\":\"https://onlinelibrary.wiley.com/doi/10.1002/ecj.12429\",\"RegionNum\":4,\"RegionCategory\":\"工程技术\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q4\",\"JCRName\":\"ENGINEERING, ELECTRICAL & ELECTRONIC\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Electronics and Communications in Japan","FirstCategoryId":"5","ListUrlMain":"https://onlinelibrary.wiley.com/doi/10.1002/ecj.12429","RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q4","JCRName":"ENGINEERING, ELECTRICAL & ELECTRONIC","Score":null,"Total":0}

引用次数: 0

摘要

研究了基于深度神经网络(DNN)的精细陶瓷烧结体扫描电镜图像分割问题。我们特别关注由晶界(粒子之间的边界)模糊引起的错误分类。因此，我们利用晶界亮度梯度的频率分布，对梯度值较低的像素点赋予较高的权重。实验证明，用所提出的损失函数训练的模型预测效果最好。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

查看原文本刊更多论文

Loss function for ambiguous boundaries for deep neural network (DNN) for image segmentation

This study deals with the task of segmentation of SEM images of fine ceramics sintered bodies by using deep neural network (DNN). In particular, we focus on misclassification caused by the blurriness of grain boundaries(boundaries between particles). Therefore, we utilize the frequency distribution of brightness gradient of grain boundaries and give higher weights to pixels with lower gradient values. Experiments confirmed that the model trained with proposed loss function gave the best prediction results.

求助全文

通过发布文献求助，成功后即可免费获取论文全文。去求助

来源期刊

Electronics and Communications in Japan 工程技术-工程：电子与电气

CiteScore

0.60

自引率

0.00%

发文量

审稿时长

6-12 weeks

期刊介绍： Electronics and Communications in Japan (ECJ) publishes papers translated from the Transactions of the Institute of Electrical Engineers of Japan 12 times per year as an official journal of the Institute of Electrical Engineers of Japan (IEEJ). ECJ aims to provide world-class researches in highly diverse and sophisticated areas of Electrical and Electronic Engineering as well as in related disciplines with emphasis on electronic circuits, controls and communications. ECJ focuses on the following fields: - Electronic theory and circuits, - Control theory, - Communications, - Cryptography, - Biomedical fields, - Surveillance, - Robotics, - Sensors and actuators, - Micromachines, - Image analysis and signal analysis, - New materials. For works related to the science, technology, and applications of electric power, please refer to the sister journal Electrical Engineering in Japan (EEJ).