V2.0 has reworked dataset, containing about 3x as many images, and ai upscaled images were removed.