It contains about 60,000+ images. The model was trained using about 12,000+ images and currently the Box skill finds similar galaxies from these 12K images.
There is metadata associated with each of the galaxies, which might be interesting to a researcher.