I don't think there is any work on that yet, but if the model is known it should be possible to derive the probability that a particular image is the output from it.