Gold Nanoparticle HRTEM Image Data Set

This is a set of 4000 unique gold (Au) nanoparticle simulated HRTEM images, for use in data-driven studies. Original Cartesian coordinates can be found from: https://doi.org/10.25919/5d395ef9a4291, corresponding to the Nanoparticle_id. 3000 nanoparticles were randomly chosen for the training set, and the remaining 1000 for the testing set. All files are in PNG format, and the naming convention is defined in the accompanying csv files that lists all of the manually extracted target labels (previously features of the original Cartesian set). Sizes range from 236 atoms to 14277 atoms, with both crystalline and non-crystalline configurations and regions. Each nanoparticle has been (randomly) rotated 340 times around the x, y and z axis (roll, pitch and yaw) after splitting, and the HRTEM simulated with abTEM software based on the multislice method. The training set contains 1,020,000 unique images, and the testing set contains 340,000 unique images. The HRTEM imaging parameters included accelerating voltage of 300kV, a divergence angle (semi-angle) of 100 mrad, a focal spread of 30 Angstroms, and a spherical aberration of 7μm with a Scherzer defocus using 0.05 for sampling.