X-ray images for X-ray testing and Computer Vision

As a service to the X-ray testing and Computer Vision communities, we collected more than 19.400 X-ray images for the development, testing and evaluation of image analysis and computer vision algorithms. The images are organized in this public database called GDXray: The GRIMA X-ray database (GRIMA is the name of our Machine Intelligence Group at the Department of Computer Science of the Pontificia Universidad Catolica de Chile). The X-ray images included in GDXray can be used free of charge, for research and educational purposes only. Redistribution and commercial use is prohibited. Any researcher reporting results which use this database should acknowledge the GDXray database by citing:

Mery, D.; Riffo, V.; Zscherpel, U.; Mondragón, G.; Lillo, I.; Zuccar, I.; Lobel, H.; Carrasco, M. (2015): GDXray: The database of X-ray images for nondestructive testing. Journal of Nondestructive Evaluation, 34.4:1-12. [ PDF ]

GDXray includes five groups of images: Castings,  Welds*, BaggagesNature and Settings. Each group has several series, and each series several X-ray images. For instance, series C0001 contains 72 X-ray images of an aluminum casting (wheel).

* The X-ray images of group Welds are included in GDXray thanks to the collaboration of Uwe Zscherpel at  Institute for Materials Research and Testing (BAM), Berlin.


There are some series with annotations. There are three kind of annotations: a) horizontal rectangles, b) oriented rectangles, and c) binary images, as follows.

wheel_annotation.png    baggage_annotation.png    weld_annotation.png


See details in readme.txt file of each series.

Examples of annotations:

Example Series Annotations Type
Casting defects C0002 ground_truth.txt a
Casting defects C0021 ground_truth.txt a
Wedling defects W0001 W0002 c
Welding defects W0001 ground_truth.txt a
Razor blades B0004 razor_blades.txt a
Razor blades B0048 ground_truth.txt b
Shuriken B0047 ground_truth.txt b
Handguns B0046 ground_truth.txt b
Fishbones N0012 N0013 c


There is a series of a checkerboard in order to calibrate a computer vision system (S0001):



In order to model the non-linear distortion of an image intensifier, series S0002 can be used.

In addition, in series B0044 there are 178 X-ray images that are full calibrated. The bag was rotated around its vertical axis in 2 degrees, and for each image there is a calibration 3x4 projection matrix. The projection matrices are stored in a MAT file called Pmatrices.mat.