EUVIMEDOpdag EUVIMED
Crossref

Deep Learning Approach for Multi-Class Segmentation in Industrial CT-Data

Tim Schanz, Robin Tenscher-Philipp, Fabian Marschall, Martin Simon

Research and Review Journal of Nondestructive Testing · 2023

DOI: 10.58286/28077 ↗

At a glance

Journal article in Research and Review Journal of Nondestructive Testing, 2023. Written by Tim Schanz and 3 co-authors.

AbstractBibliografiske oplysningerCiteRelated studies

Abstract (original language)

In this work, a multiclass segmentation of defects on industrial CT data is performed. The segmentation is implemented by using artificial neural networks (ANNs). Different architectures are evaluated and compared to find the best model architecture for this task. Different metrics for segmentation and classification analysis are used to evaluate the networks. The three best models are presented in this paper. All of them obtained positive results; the model with the U-Net architecture achieved the best results with a remarkable segmentation performance of 90.18% and a classification performance of 95.77%.

Cite

APA

Schanz, T., Tenscher-Philipp, R., Marschall, F., & Simon, M. (2023). Deep Learning Approach for Multi-Class Segmentation in Industrial CT-Data. Research and Review Journal of Nondestructive Testing. https://doi.org/10.58286/28077

Vancouver

Schanz T, Tenscher-Philipp R, Marschall F, Simon M. Deep Learning Approach for Multi-Class Segmentation in Industrial CT-Data. Research and Review Journal of Nondestructive Testing. 2023. doi:10.58286/28077

BibTeX
@article{Schanz2023,
  author = {Schanz, Tim and Tenscher-Philipp, Robin and Marschall, Fabian and Simon, Martin},
  title = {Deep Learning Approach for Multi-Class Segmentation in Industrial CT-Data},
  journal = {Research and Review Journal of Nondestructive Testing},
  year = {2023},
  doi = {10.58286/28077}
}
Start for free →