Identified Fibers and Validation Data for Fiber Identification

in an ultra-large µCT scan of a nonwoven sample

The following data sets belong to the µCT scan of a nonwoven sample provided by Reifenhäuser Reicofil GmbH & Co. KG.

Aside from the actual scan, in the first data set, we provide a segmentation of the scan into pore space and fibers.

In addition, we supply another virtual 3D scan in which the individual fibers are identified as separate objects. This identification was carried out using a Neural Network validated with artificial data, which make up the second data set.

The methodology is part of the doctoral thesis of Andreas Griesser and it is described in more detail in an upcoming article. The size of the nonwoven sample scan and the virtual scan with the identified fibers is 15,363 x 3,960 x 2,112 voxels each. The ultra-large size is intended and suitable as a benchmark for real nonwovens and high-performance algorithms.

The data has been used by Reifenhäuser Reicofil GmbH & Co. KG in their Patent No. WO 2020/103964 A1. The algorithms were thereby required for the proof of patentability of the nonwoven.

Griesser A., Westerteiger R., Glatt E., De Boever W., Hagen H., and Wiegmann A.: Identification and analysis of fibers in ultra-large micro-CT scans of nonwoven using Deep Learning, accepted with modifications in the Journal of the Textile Institute. 

1. Dataset with µCT scan and results from fiber identification in a nonwoven sample

The downloadable dataset contains all scan data. The data collection, image processing, and analysis of the nonwoven sample metadata are explained in the data description.

Download dataset    Data description

The size of the ZIP file is 4.5 GB. The uncompressed data is about 720 GB.

Citation of dataset:

Griesser A., Westerteiger R., Glatt E., De Boever W., Hagen H., and Wiegmann A., 2022: SampleC - micro-CT and fiber identification of a nonwoven sample, Math2Market GmbH, Sample-C, ​​https://doi.org/10.30423/Data.Math2Market-2022-02.Sample-C.FiberFind

Citation of article:

Griesser A., Westerteiger R., Glatt E., De Boever W., Hagen H., and Wiegmann A.: Identification and analysis of fibers in ultra-large micro-CT scans of nonwoven using Deep Learning, accepted with modifications in the Journal of the Textile Institute. 

2. Dataset with validation of fiber identification - Basis and results of fiber identification for generated samples

The dataset consists of 3 samples, each containing 8 different files. The base files are converted outputs from the FiberGeo (fiber generator) module of GeoDict. The result files are converted outputs from the FiberFind (fiber identification) module of GeoDict.

Download dataset    Data description

The size of the ZIP file is 163.4 MB. The uncompressed data is about 2.8 GB.

Citation of dataset:

Griesser A., Westerteiger R., Glatt E., Hagen H., and Wiegmann A., 2022: Fiber identification validation - ground truth and results of fiber identification for generated samples, Math2Market GmbH, Validation, ​​https://doi.org/10.30423/Data.Math2Market-2022-02.Validation.FiberFind

Citation of article:

Griesser A., Westerteiger R., Glatt E., De Boever W., Hagen H., and Wiegmann A.: Identification and analysis of fibers in ultra-large micro-CT scans of nonwoven using Deep Learning, accepted with modifications in the Journal of the Textile Institute 

Authors

Andreas Griesser   (Math2Market GmbH, Germany),
Dr. Rolf Westerteiger   (Math2Market GmbH, Germany)

Collaborators
Dr. Erik Glatt   (Math2Market GmbH, Germany),
Wesley De Boever, PhD  (at that time Bruker, Belgium),
Prof. Dr. Hans Hagen  (TU Kaiserslautern, Germany)
Andreas Wiegmann, PhD   (Math2Market GmbH, Germany)

Published: August 2022
License: ODC-BY 1.0, https://opendatacommons.org/licenses/by/1-0/