Neural networks, upscaling, denoising

Pages: 1

30.05.2018 19:24:00
On the left is the original, which was reduced to 2,104 by the short, then increased 4 times, the bicubic in the middle, the neural network to the right.
comparsion. jpg on ixbt. photo:


All originals in the album http: // ixbt. photo /? id = album: 61650
The site that does this: https: // letsenhance. io / If you know cheaper / free / locally / any alternatives, throw


Forever branches " when the megapixel race stops " - greetings
 

30.05.2018 21:06:00
chivo, chivo?

30.05.2018 21:18:00
Is it clean, chivo?
Neural networks allow you to magnify images by tracing them and many more.
https: // www. youtube. com / watch? v = OSShk6jA_us
most meat begins with 8: 38

That Noise
https: // www. youtube. com / watch? time_continue = 2 & amp; v = qWKUFK7MWvg

30.05.2018 21:24:00
but, it's justified that the killer did not find

31.05.2018 5:33:00

Left of the original, which was reduced to 2,104 in the short, then increased 4 times, the bicubic in the middle, the neural network to the right.
Has it been reduced by something-a cube or a neuron?

31.05.2018 8:05:00
Apskale is of little interest. More interesting is the demosuke.

31.05.2018 11:09:00
:

Left original, which was reduced to 2104 in the short, then increased 4 times, the bicubic in the middle, the neural network to the right.
And was it reduced by something-a cube or a neuron? In short, from here this is
small. jpg on ixbt. photo:


she did this is:

neural. jpg on ixbt. photo:
:
Apskale is of little interest.
A breakthrough in the area of ​​upscale such as never before in history - " Apskale is of little interest " ?
 

31.05.2018 11:41:00

A breakthrough in the area of ​​upscale such as never before in the history of
What? Details, of course, do not appear (comparison. Jpg), but do square, to contrast the sigma for a long time knows how.
Unless, as the technology of increasing the local contrast - increase the network, then reduce it back.

" Apskale is of little interest " ?
Apskeylit is not necessary from the word at all, but it is necessary to deal with a demosuke continuously.

31.05.2018 11:44:00
:

Breakthrough in apskeyla such, which have never in the history has not been
What? Details, of course, do not appear (comparison. Jpg), but do square, to contrast the sigma for a long time knows how. .
Well, show me this, if you can. Take small. jpg and make it 8416 x 7360. Yes, it's soapy, but nobody really needs this increase, it's time, but here's a little bit to increase or remove blurring - the potential is huge. And the neural network itself is trained, that is, the further it will be, the faster it will be.

31.05.2018 12:00:00
I tried to increase the proposed ones and reduce them back. For some reason, the contrast of shadow / light changed. And brought the blues.
In the enlarged image, the details did not appear, of course, the artifacts of the demozaic were stressed, and there was also no underlined clarity of details.
But, if you bring sharps to the reduction, you get clarity, which is not easy to get right in the original image.


Well, show me this, if you can.
Do I look like a sigma engineer? I compared the results of the treatment with what lies in RAW sigma.


And the neural network itself learns
Where does it learn there?
You can teach it by sending a large original and a thumbnail copy " Look, from such a small one it should turn out such a big " . And there is no such service I do not observe.


31.05.2018 12:42:00

Well, show me if you can.
Do I look like a sigma engineer? I compared the results of the treatment with what lies in RAW sigma. Comparing the airplane and the discriminant.

and neural network trained herself
At what point is she studying?
You can train it by sending a large original and a small copy of the " Look, from such a small one it should turn out such a big " . And there is no such service I do not observe.
The fact that a specific network on a specific site users are not allowed to teach - does not mean that it is impossible in principle.
 

04.10.2018 20:31:00
Pages: 1

Neural networks, upscaling, denoising

info@www.about-digital-photo.com