The learning happens on the level of seeing what would a human retoucher does by using the data we gathered from the editor: https://www.polarr.co/editor