tanks bakes!
>They just lost the trust of the majority of those who watch it.One America News Networkis going to have high ratings next month.
/fixed/
o7
>If it was any kind of comm at all, the comm is "Look at Disney" not the time.
this anon thinks possible some [DS] faggot owns one of these and has on his office/ bedroom wall or some shit
so, in muh theory
anon was talking to some faggot he knew lurks the boards, and was basically saying;
all your amazon echo are belong to us
>>10690981Notable
>meme scrambler
anons check out?
report back results?
what does the output look like?
>another another said they did check it. it's in notables.
5:5 bakes
anon also reposted in meme camo thread here:
(in global notables atm)
thanks for keeping eyes on
o7
>All got identified as the mickey clock
dots are probably too small
sniffer can still "see" the architecture of the orriginal image
try bigger squares
these have been tested and work
>Throw a mostly transparent image of a crosswalk over a meme, and algo will see a crosswalk.
^^^^^THIS^^^^^
1000x this
The human eye see's the meme (if you do the camo right)
The AI sees the transparency layer (again, if you do it right)
It is the fact that the AI is seeing your transparency layer that is causing the mitigation
If the AI didnt see the transparency layer, it would be seeing whats left, the original meme
That Being Said
just about any pattern is going to work, provided the block size of the pattern is large enough
it has been tested that small dots interference actually doesnt work, the algo still sees the underlying meme
pic related
SO
Why dont we use official Democrati Party logos and election media as our transparency layers?
The AI will start identifying the transparency layer in order to counter our meme insertion.
As the AI "learns" the transparency layer it needs to filter, it will be "learning" official democratic party media logos
hence, theoretically, it will pop false positives on genuine Democratic media
poison the filter
all your AI are belong to us
o7
forgot the pic related
as you can see
this is an example of aFAIL
I actually thought this would pass
but, as you can see…
o7
>all three got sniffed
this is all through tineye right?
absolutely should develop the most robust techniques for beating image snifers
but
it seems Twitters image sniffer is behind tin-eye
because all of those techniques beat Twitters sniffer
-keks
>This didn't come up with a result, with Yandex
5:5
but original subjact matter integrity has been…. sacrificed severely
/obviousanon is obvious
keep working it, if we figure out how to beat the tineye sniffer, Twitter/FB wont have a chance, even as they try to deply countermeasures