5 images crop of the face, all from the famous painting (3 using codeformer at 95% fidelity to upscale, since all images are the same but only on slightly different lighting, quality and colors, background was blurred/altered deliberately on 3 or 4 of the images so that the training would focus on the face, then trained at batch 2 and 3 of the other stuff to about 2200 but realized it was garbage after 1700 decided to seletect 475 (550 was decent too) to upload