Ground truth and CNN model prediction images Initial dataset size: 1703 non-linear dynamic (time history) simulations. Input parameters include undeformed steel frame images providing geometry characteristics such as bay and storey numbers, as well as thicknesses of steel plate shear walls, fundamental frequencies of each building and peak ground accelerations of the seismic loading events. Output parameters are images providing failure on steel shear walls, in the form of accumulated plastic distribution. A dataset of 1703 pair of images (groundtruth and predicted) have been generated, providing a range of structural and loading parameters for multi-storey steel buildings with SPSWs. Non-linear dynamic finite element simulations have been performed, using as input parameters the number of baysalong the two horizontal directions, the number of storeys, the thickness of the shear plate walls and the peak ground acceleration. Output (predicted) parameters are images depicting failure on the shear plate walls, in the form of equivalent plastic strain distribution. In addition to the images, each building is associated with a set of metadata describing the structural and seismic parameters. These include Length, width, and height: measured in number of bays along the two horizontal directions and height of each storey: wall thickness: structural thickness of each SPSW PGA (Peak Ground Accelerator): a measure of earth-quake intensity POV (point of view): the view direction, from one of the four sides (A,B,C or D) Hz: The dominant frequency of the ground motion