Dataset for Popov, Strokov, and Surdyaev 2021: Black hole shadows against an optically thick, geometrically thin disk

The datasets contain images of an optically thick, geometrically thin accretion disk distorted by the presence of a Kerr black hole (i.e., black hole shadows). The images are in a compressed text format (.gz) as well as in a graphic format. The training dataset comprises 2,301 images for various values of the disk's outer radius, the spin and sense of rotation of the black hole, and for different viewing angles (w.r.t. the equatorial plane where the disk resides). The test set contains 89 images with parameters which were randomly sampled from the same ranges as in the training set. For details, see the works associated with this record.