FOD CT Data: air pockets in avocado and stone in modelling clay

Summary

This submission contains X-ray CT data of avocado fruits and pieces of modelling clay containing pebble stones.
Data for every object include binned pre-processed projections and volume segmentations.
These datasets can be used for training and testing deep learning methods for foreign object detection.

The data is made available as a part of the paper "CT-based data generation for foreign object detection on a single X-ray projection".

Data acquisition

A majority of raw data for modeling clay (excluding 10 samples without pebble stones in the Test subset) is taken from the dataset
"A collection of 131 CT datasets of pieces of modeling clay containing stones"
[](https://doi.org/10.5281/zenodo.5866228)

The remaining pieces of modeling clay and all avocado fruits were scanned at the FleX-ray laboratory
of the Centrum Wiskunde & Informatica (CWI) in Amsterdam, the Netherlands (details can be found in [Coban 2020]).
For every fruit, we made scans with significantly different amounts of air pockets by waiting for a few days between experimental acquisitions.
The measurements were performed with the voltage of 90 kV, power of 45 W, exposure time of 300 ms per projection, and magnification factor of 1.3.
The original X-ray image size was 1912 px x 1520 px with a pixel size of 75 μm, 1440 images were acquired for every sample.
For faster deep learning model training, images and reconstructions were downsampled with a factor of 4, leading to the effective pixel size of 300 μm and voxel size of 230 μm.
Additional scans of the pieces of modeling clay were acquired with settings similar to the main collection.

Data Description

The submission is split into "Avocado" and "Playdoh" (pieces of modeling clay) datasets. Each dataset is further split into Training and Test subsets.

The folder for every scanned object contains
- ./log/ - subfolder with logarithmed X-ray projections after darkfield and flatfield correction.
- ./segm/ - subfolder with slices of the segmented volume.
- ./scan settings.txt - a file with scanner metadata containing scan geometry
- ./volume_info.csv - a file with a voxel count for every class in the segmentation.

For playdoh objects, the segmentation classes are modeling clay (Class 1) and pebble stone (Class 2). In this case, pebble stones are foreign objects.

For avocado objects, the segmentation classes are peel (Class 1), avocado meat (Class 2), seed (Class 3) and air pockets (Class 4). Air pockets are considered a foreign object.

Additional Links

These datasets are produced by the Computational Imaging group at Centrum Wiskunde & Informatica (CI-CWI). For any relevant 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 datasets, please get in touch with
- vladyslav.andriiashen [at] cwi.nl

References

[Coban 2020] S. B. Coban, F. Lucka, W. J. Palenstijn, D. Van Loo, and K. J. Batenburg, “Explorative imaging and its implementation at the FleX-ray Laboratory,” J. Imaging, vol. 6, no. 18, 2020, doi: 10.3390/jimaging6040018.