{"title":"Enhancing Semantic Segmentation in High-Resolution TEM Images: A Comparative Study of Batch Normalization and Instance Normalization.","authors":"Bashir Kazimi, Stefan Sandfeld","doi":"10.1093/mam/ozae093","DOIUrl":null,"url":null,"abstract":"<p><p>Integrating deep learning into image analysis for transmission electron microscopy (TEM) holds significant promise for advancing materials science and nanotechnology. Deep learning is able to enhance image quality, to automate feature detection, and to accelerate data analysis, addressing the complex nature of TEM datasets. This capability is crucial for precise and efficient characterization of details on the nano-and microscale, e.g., facilitating more accurate and high-throughput analysis of nanoparticle structures. This study investigates the influence of batch normalization (BN) and instance normalization (IN) on the performance of deep learning models for semantic segmentation of high-resolution TEM images. Using U-Net and ResNet architectures, we trained models on two different datasets. Our results demonstrate that IN consistently outperforms BN, yielding higher Dice scores and Intersection over Union metrics. These findings underscore the necessity of selecting appropriate normalization methods to maximize the performance of deep learning models applied to TEM images.</p>","PeriodicalId":18625,"journal":{"name":"Microscopy and Microanalysis","volume":" ","pages":""},"PeriodicalIF":2.9000,"publicationDate":"2025-02-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Microscopy and Microanalysis","FirstCategoryId":"5","ListUrlMain":"https://doi.org/10.1093/mam/ozae093","RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"MATERIALS SCIENCE, MULTIDISCIPLINARY","Score":null,"Total":0}
引用次数: 0
Abstract
Integrating deep learning into image analysis for transmission electron microscopy (TEM) holds significant promise for advancing materials science and nanotechnology. Deep learning is able to enhance image quality, to automate feature detection, and to accelerate data analysis, addressing the complex nature of TEM datasets. This capability is crucial for precise and efficient characterization of details on the nano-and microscale, e.g., facilitating more accurate and high-throughput analysis of nanoparticle structures. This study investigates the influence of batch normalization (BN) and instance normalization (IN) on the performance of deep learning models for semantic segmentation of high-resolution TEM images. Using U-Net and ResNet architectures, we trained models on two different datasets. Our results demonstrate that IN consistently outperforms BN, yielding higher Dice scores and Intersection over Union metrics. These findings underscore the necessity of selecting appropriate normalization methods to maximize the performance of deep learning models applied to TEM images.
将深度学习整合到透射电子显微镜(TEM)图像分析中,对推动材料科学和纳米技术的发展大有裨益。深度学习能够提高图像质量,实现特征检测自动化,并加速数据分析,从而解决 TEM 数据集的复杂性问题。这种能力对于精确、高效地表征纳米和微米尺度的细节至关重要,例如,有助于对纳米粒子结构进行更准确、更高通量的分析。本研究探讨了批量归一化(BN)和实例归一化(IN)对高分辨率 TEM 图像语义分割深度学习模型性能的影响。利用 U-Net 和 ResNet 架构,我们在两个不同的数据集上训练了模型。我们的结果表明,IN 的性能始终优于 BN,其 Dice 分数和交集指标均高于联合指标。这些发现强调了选择适当归一化方法的必要性,以最大限度地提高应用于 TEM 图像的深度学习模型的性能。
期刊介绍:
Microscopy and Microanalysis publishes original research papers in the fields of microscopy, imaging, and compositional analysis. This distinguished international forum is intended for microscopists in both biology and materials science. The journal provides significant articles that describe new and existing techniques and instrumentation, as well as the applications of these to the imaging and analysis of microstructure. Microscopy and Microanalysis also includes review articles, letters to the editor, and book reviews.