This submission contains supplementary material to A.A. Hendriksen et al., (2019). Throughout the description, we mention this publication as the "referred paper".

Description Summary: "Oatmeal Data" is a collection of experimental tomographic data of a plastic jar filled with store-bought mixed-size flake oatmeal. The dataset was collected to test methods specifically developed for improving image resolution in a reconstructed volume (primarily for machine learning). As well as the raw (unprocessed) tomographic data, this submission includes training and test data, weights and geometry parameters, and a text file (Makefile) to reproduce the results presented in the referred paper. The full list of contents is given below.

Apparatus: The dataset is acquired using the custom-built and highly flexible CT scanner, FleX-ray Laboratory, developed by XRE NV and located at CWI. This apparatus consists of a cone-beam microfocus X-ray point source that projects polychromatic X-rays onto a 1944-by-1536 pixels, 14-bit, flat detector panel.

Sample Information: The sample consists of a thin-walled plastic far filled with store-bought mixed-sized oatmeal flakes. This sample was chosen due to its repeating/self-similar structure. This self-similarity is an important factor and is exploited in the algorithm for improving image resolution, presented in the referred paper.

Experimental Plan: Three experiments at two magnification levels were performed, which are denoted by Zoom1 and Zoom4: Zoom1 is the dataset collected with magnification factor of 1.09, with the plastic jar filled with oatmeal fully in view. Zoom4 - centre is the sample scanned at magnification of 4.38, with the field of view aligned with the centre of the same (i.e. the centre of the original view in Zoom1). Zoom4 - top is where the tube and the detector are positioned to show the top of the sample (surface level of the oatmeal flakes is visible at the top of the field of view). The datasets were collected over a 360° in circular and continuous motion with 2000 projections distributed evenly over the full circle. The upload includes the raw data collected following the experimental plan, which includes additional projections for normalization: one dark-field (closed shutter) and two (pre- and post-) flat-field (open shutter) images. Each dataset is packaged with the full list of data and scan settings files (in .txt format). These files contain the tube settings, scan geometry and full list of motor positions at the start of the scan. In addition, ASTRA geometry format file is included for each data in the geometries folder.

List of Contents: The contents of the submission (sorted into folders) is given below. data: The raw (uncorrected) datasets (including dark and flat fields) for Zoom1, Zoom4-centre, Zoom4-top. geometries: The set of parameters needed for the ASTRA forward and back projectors. test: Test dataset compiled as detailed in the referred paper, or in On-the-fly GitHub documentation. train: Training dataset as detailed in the referred paper, or in On-the-fly GitHub documentation. weights: Weights as detailed in the referred paper, or in On-the-fly GitHub documentation. Makefile: File containing steps to reproduce results presented in the referred paper.

Additional Links: These datasets are produced by the Computational Imaging group at Centrum Wiskunde & Informatica (CI-CWI). For any useful Python/MATLAB scripts for the FleX-ray datasets, we refer the reader to our group's GitHub page.

Contact Details: For more information or guidance in using these dataset, please get in touch with s.b.coban [at] cwi.nl allard.hendriksen [at] cwi.nl

, , ,
doi.org/10.5281/zenodo.2657644
creativecommons.org/licenses/by/4.0/legalcode
Computational Imaging

Coban, S.B, Hendriksen, A.A, Pelt, D.M, Palenstijn, W.J, & Batenburg, K.J. (2018). Oatmeal Data: Experimental cone-beam tomographic data for techniques to improve image resolution. doi:10.5281/zenodo.2657644