The first article about "Generative Adversarial Networks and De-Mosaicing" covers a different form of "demosaicing". It can be read here: https://en.wikipedia.org/wiki/Demosaicing . This is a different problem than pixelated faces/genitals.
The third article about "Google De-Mosaicing Faces" has some important parts:
"When some details do not exist in the source image, the challenge lies not only in “deblurring” an image, but also in generating new image details that appear plausible to a human observer."
"The published images shown by Google Brain’s neural system bear a resemblance to the real person or scene but not necessarily enough to serve as anything better than an approximation."
"But getting to grips with real faces at awkward angles depends on numerous small details. Emphasise the wrong ones and police could end up looking for the wrong person."
"In the end, this kind of approach is probabilistic – another way of saying it’s a prediction. In the real world, predictions sound better than nothing, but a more reliable end for CCTV might be achieved simply by improving the pixel density of security cameras."
The last paragraph pretty much describes what neural nets are doing. Neural networks learn training sets and with a certain input they guess an output based on what they have learned. In this case here they guess how to "remove" the pixelation of genitals based on the uncensored genitals examples they have learned. A "reconstructed" vagina could be an average vagina of certain vaginas in the training set. In other words, you will see fake genitals. Imho, this is not the answer to the pixelation of the genitals in JAVs.
Another thought on the algorithms for the pixelation of the genitals: Mathematically, the used pixelation algorithms represent functions that aren't bijective and therefore these functions have no inverse. This is why it is impossible to find an algorithm that simply undoes the pixelation. In other words, the algorithms for the pixelation destroy information and therefore it can't be undone. That is also the reason why a neural network won't learn/find a function that undoes the pixelation algorithm. A neural network can only guess a solution based on the training set. Any other solution would also create fakes.
