Quote:
Originally Posted by zikzak
I'm going to go out on a limb and guess that tech nerd stories about photo enhancement becoming a real thing are 95% misleading clickbait bull****, and the 5% that is true is obvious and unsurprising.
It is called super-resolution. Then combined with a facial recognition database as every major corporation and government agency possesses, it can be used to great effect.
The first column is the input into the network. The third column what the network produces as output. The fourth column is the ground truth.
Quote:
Originally Posted by Noodle Wazlib
I guess it doesn't count as not there when it turns out it is, but didn't some MIT folk find out a way to extract sound from still images or something crazy like a year ago?
It was recovering sound from video without sound, which is pretty awesome. This was also done in 2014, before many advances in deep neural networks.