|
|
|
|
|
|
by d--b
3286 days ago
|
|
|
That's the point of using deep learning here. Of course you can't make up the missing information, but by training the model with a lot of samples, it should eventually reach a point where it produces the most likely original information. It works quite well on images: https://github.com/alexjc/neural-enhance |
|
|