Dataset of Synthetic X-ray Scattering Images for Classification Using Deep Learning

This dataset contains a large number of example x-ray scattering images; each image is tagged with a variety of attributes describing the data features appearing in the image ('rings', 'anisotropic', etc.) or describing the underlying material ('BCC', 'FCC', etc.). The main purpose of this dataset is as a training set for machine-learning methods. The images were generated synthetically, using a combination of ad hoc methods (e.g. superimposing features such as rings and halos) and simple simulations (e.g. generating realspace arrangements of nanoparticles, and then computing the far-field scattering pattern). The presented code iterates across a wide variety of input conditions, such that the output images cover a wide range of expected x-ray scattering image types. Experimentally-realistic artifacts, including masks, parasitic streaks, and Poisson noise, are also included.