Find research datasets worth reusing

Unpaired fast- and slow-acquisition microCT scans of carbon-fibre-reinforced composites

<p>data.zip contains tomographic images of carbon-fibre-reinforced composites with <a href="https://www.toraycma.com/wp-content/uploads/T700S-Technical-Data-Sheet-1.pdf.pdf">Toray T700S</a> fibres in a <a href="http://sicomin.com/datasheets/product-pdf1123.pdf">Sicomin SR8500 KTA313</a> epoxy matrix.</p> <p>The datasets accompany the paper <a href="https://doi-org.remotexs.ntu.edu.sg/10.1016/j.compscitech.2023.110278" target="_blank" rel="noopener"><em>Deep-learning image enhancement and fibre segmentation from time-resolved computed tomography of fibre-reinforced composites</em></a>.</p> <p>They are used in a two-step process:<br>1) to train a 3DCycleGAN enhancement algorithm available on <a href="https://github.com/pvilla/3DCycleGaN">https://github.com/pvilla/3DCycleGaN</a>.<br>2) to train a UNetID segmentation algorithm available on <a href="https://github.com/CMG-KULeuven/UnetID-FibreSeg">https://github.com/CMG-KULeuven/UnetID-FibreSeg</a>.</p> <p>The tomograms were acquired at the TOMCAT Beamline at the Swiss Light Source in October 2021.</p> <p>Acquisition specifications:</p> <p>T700_GF_0p8um_1ms:</p> <ul> <li>detector: GigaFRoST</li> <li>effective voxel size: 800x800x800 nm&sup3;</li> <li># projections: 500</li> <li>exposure time: 1 ms</li> <li>total acquisition time: 500 x 1 ms = 0.5 s</li> <li>beam: polychromatic</li> <li>datasets: (same sample, different regions)<br>T700-T-21_GF_0p8um_1ms_1.h5<br>T700-T-21_GF_0p8um_1ms_2.h5<br>T700-T-21_GF_0p8um_1ms_3.h5</li> </ul> <p>T700_GF_1p6um_0p5ms:</p> <ul> <li>detector: GigaFRoST</li> <li>effective voxel size: 1600x1600x1600 nm&sup3;</li> <li># projections: 500</li> <li>exposure time: 0.5 ms</li> <li>total acquisition time: 500 x 0.5 ms = 0.25 s</li> <li>beam: polychromatic</li> <li>files: (same sample, different regions)<br>T700-T-21_GF_1p6um_0p5ms_1.h5<br>T700-T-21_GF_1p6um_0p5ms_2.h5</li> </ul> <p>T700_GF_1p6um_3ms:</p> <ul> <li>detector: GigaFRoST</li> <li>effective voxel size: 1600x1600x1600 nm&sup3;</li> <li># projections: 500</li> <li>exposure time: 3 ms</li> <li>total acquisition time: 500 x 3 ms = 1.5 s</li> <li>beam: polychromatic</li> <li>files: (same sample, different regions)<br>T700-T-21_GF_1p6um_3ms_1.h5<br>T700-T-21_GF_1p6um_3ms_2.h5<br>T700-T-21_GF_1p6um_3ms_3.h5<br>T700-T-21_GF_1p6um_3ms_4.h5<br>T700-T-21_GF_1p6um_3ms_5.h5</li> </ul> <p>T700_pco_0p4um:</p> <ul> <li>detector: pco.edge 5.5</li> <li>effective voxel size: 400x400x400 nm&sup3; (scaled from 325x325x325 nm&sup3;)</li> <li># projections: 2000</li> <li>exposure time: 250 ms</li> <li>total acquisition time: 2000 x 250 ms = 500 s</li> <li>beam: 15 keV</li> <li>files: (different samples)<br>T700-T-02_pco_0p4um.h5<br>T700-T-08_pco_0p4um.h5<br>T700-T-21_pco_0p4um_reference.h5<br>T700-T-26_pco_0p4um.h5</li> </ul> <p>3DCycleGAN_trained_models.zip contains pre-trained models for the 3DCycleGAN enhancement algorithm.</p> <p>UnetID_trained_models.zip contains pre-trained models for the UNetID segmentation algorithm.</p>