Longsworduser reliable description :

https://www.pixiv.net/member_illust.php?mode=medium&illust_id=69781026
https://ec.toranoana.jp/tora_r/ec/item/040030649905/
https://www.melonbooks.co.jp/detail/detail.php?product_id=377697
https://www.doujinshi.org/book/1588613/

Decensored with :
https://github.com/deeppomf/DeepCreamPy

射爆了，谢谢茄子

Base +2, 123ecq347 +6, ShirokiinChitoge +6, fuck1712938443 +4, 夜神诗 +3, terryproprieter +6, [email protected] +2

not using the original resolution?

Base +6, Jin.H. +8, Ancient-Alien +11, SocksGalore +6, dj756 +6, Tenacia_Ferrari +6, Oniichananon +6, SKMMY -30, TheOldDeus +7, Lyncs +11, Lostalgia +8, timbus +11, IonicTHI +7, and 2 more...

>Decensored with :
>https://github.com/deeppomf/DeepCreamPy

The future is now

Base +7, Jin.H. -8, sneikkimies -23, SocksGalore +6, masum_beast +4, dj756 -6, Ckraemer624 -4, riseed55 +4, Fomkin +5, BismarckDeri +6, aubrie paterson +6, Putty97525 +6, platewolf +9, and 32 more...

With some color leveling this decensoring would be perfect... guess we can declare white-stripe censoring is cracked... Will this work on mosaic too?

Base +7, SocksGalore +6, dj756 +6, Quartz Saint +4, Tenacia_Ferrari +6, unicorn901 +7, TheOldDeus +7, Lostalgia +8, amurec80 +6, arbitraryalias +6, Sdoots +6, shuntensatsu +15, enterthename +6

Deep learning for hentai.

Base +17, unicorn901 +7, TheOldDeus +7, dar4545 +9, Nep17 +5, amurec80 +6, bunnygunny +6, cryan_g +5, Sdoots +6, shuntensatsu +15, Megaton +39, jpabad +6, enterthename +6, and 9 more...

what the difference beetwen Decensored and uncencored?

Base +6, TheOldDeus +7

Yes it also work with mosaic.

Farmuhan, the fact that you don't see the difference make me proud of the work. Uncensored is more when the artist give the version of his work whithout the censor, descensored it's when a third party came from the censored version and try to make an uncensored version of it.

Now i upload it, i see some mistake (green on 10 and 18), i'll correct them as soon as possible. After that i upload the next project.

Base +1, TheOldDeus +7, my little donger +7, Sdoots +6, num21 +6, illusioninusall +16, Luc3at3i5 +7

Can you give a description of how this works? This is fascinating.

Base +9, TheOldDeus +7, amurec80 +6

Yeah of course, i've not done the big part of the work : the github https://github.com/deeppomf/DeepCreamPy
You have to install it (in my case i'm on linux and it seem to be more easy to install on it).
After that i use an image editing program to color the censored part in green (only green in hexa 00FF00 work).
I put the picture in the input folder and execute the python code (it see the green part and descensored it).
It done a really good job but the fact is that you had to rework on the image if you want a great thing !

I hope I made you want to try, it's pretty easy and very funny.
If you want help like I said in picture 23 you can contact me by twitter @poketshort or by mail "[email protected]"

Base +1, TheOldDeus +7, my little donger +7, see-eff +6

Thanks for sharing.

Base +1, TheOldDeus +7, shuntensatsu +15, HentaiJew +6

It doesn't work. I mean if you wan small disturbing rectangles somewhat removed, it's okay. But it just can't understand border between object. Even contrast one. Which, btw, is sometimes work okay in photoshop, since you can use C.A.fill in subsequence. At best moments it's +- close to content aware fill from photoshop, which BTW, recently got a big update in functionality and usability. Without it being able to analyze genitals shape and use it to guess how to fill , there is no reason to go to such length using standalone command line content aware fill, especially if the end you will need to do huge manual edits on it. I mean, unless you doesn't have access to photoshop. "neural "network"". For example there is algorythm to manually fit generic face mesh to photo in nuke by using control points with very good results, and there are ages of face recognition libraries out there, there is even couple of implementation of this two together already albeit with limited functionalyt. And recognising one cylinder and one crack shouldn' be that much harder especially if you tell a program where to look.

Last edited on 01 November 2018, 15:13.

Base +7, B1u3z +7, Hyoros -15

@depola i see, thanks for the explanation

Base +6

https://www.pixiv.net/member.php?id=853087

Base +7

Nice woooooork.

Base +5

Farmuhan was asking what the difference between the terms were, not asking how this gallery differed. Sorry to say, but you have to have pretty bad eyesight to not be able to tell the difference here.

Base +30, Megaton +39, neteroisgod +6, PinguPrin +6, Hyoros +15

Still missing the Abigail sex page (Chaldea_Maid_003.jpg), if it ever existed, that is.

Base +4, Elian_JMP +3, see-eff +6, Luc3at3i5 +7, nasilacus +6

这无修正是自己改的呀, 并非作者的原画

Base +7, galgametentacle -6, Badplayer17 +6

AI无修有待改进……

Last edited on 28 January 2019, 17:34.

Base +26