What if I now get a bunch of images before & after and train a neural network on this to "learn" the mapping. Who would the neural network belong to?