**By Hans-Henrik Fuxelius, 2023-08-21
The objective here is, by dimensionality reduction, to take an image as input and reproduce an image as output by regression. It does not learn the features of the images, but the rules for transformation to desired output!
Since the image2image transform is possible, so must imaging be (image2mirror) since it is just a reordering of the output order!
The nn learns the transform for how to copy any image from input to output
The nn learns the transform for how to mirror any image from input to output
