An ex-girlfriend of mine used to work on a newspaper's website. A very staid and respectable newspaper's website. Actually, an enormously staid and respectable newspaper's website. They had a severe problem with this method of evaluation (which was in use by nannyware companies even then, about six years ago). It meant their site was barred, in spite of their tremendously staid and respectable nature, due to the amount of fleshtone on display.
My college prof was working on something like this for our machine vision class. Every algorithm we tried failed on the same edge case: Closeups of human faces.
Lame. HUMANS can't even solve the "What's obscene?" problem.
Here's a function to tell you whether an array of bytes (with length and width provided) composes an obscene picture that has gives NO false positives!
Face detection doesn't seem too hard. Distinguishing between a face and a face with "a penis resting on it" (as Dan puts it) might be a little more difficult...
i remember a guy from my faculty, too (back in 1998). the only one with 730 mb porn on his quota (all in jpg), he called "test samples". *shrug* whenever you made a locate something | grep thisandthat - all you get was asian slu..... as bycatch. and his algorithm didn't work either (since it never occured in his vita). actually his name was dicken... also i recall a theorem from computer visons lecture which was something like: "till now there is no computer vison system with power and capabilities of a biological vision system like the human - which complexity is too high to be totally understood even by todays knowledge" blabla retina + optical chiasma vs. neural networks
I'm not good at math, so I need you guys to help me with something. Based on the information above, I'm mostly interested in those parts of the internets where V < 10. Anybody know any sites where V < 30, or V < 30n, because those parts of the internets are probably even better, right?
I wonder if this is anything like makeup or bandages, where it's only for caucasian nudity. If so, I see a steady increase in the sweet chocolatification of the internets.
Also, 246 gives me a boner.
Comments are closed because this post is 18 years old.
I bet they HATED compiling that "questionable shape database". Uh huh.
*blink*
So, if I'm looking at this correctly, an image of solid color would be detected as skin?
The worst thing about Internet pornography is how it pressures the young, impressionable models into getting those dangerous high-heel implants.
Hahah
just start making porn stars with green skin!
Oh, no! Then we'll never get rid of Shatner!
An ex-girlfriend of mine used to work on a newspaper's website. A very staid and respectable newspaper's website. Actually, an enormously staid and respectable newspaper's website. They had a severe problem with this method of evaluation (which was in use by nannyware companies even then, about six years ago). It meant their site was barred, in spite of their tremendously staid and respectable nature, due to the amount of fleshtone on display.
This was, of course, the Financial Times.
Woah, this is why. CHILD PRON on the front page!!1
My college prof was working on something like this for our machine vision class. Every algorithm we tried failed on the same edge case: Closeups of human faces.
dumb, dumb, dumb.
Lame. HUMANS can't even solve the "What's obscene?" problem.
Here's a function to tell you whether an array of bytes (with length and width provided) composes an obscene picture that has gives NO false positives!
boolean isObscene(Image p0rn0) { return false; }
Plz to be sending my comp. sci. doctorate now.
Whoops, that function didn't quite meet the spec listed.
I guess that's why we have debugging tools!
Face detection doesn't seem too hard. Distinguishing between a face and a face with "a penis resting on it" (as Dan puts it) might be a little more difficult...
i remember a guy from my faculty, too (back in 1998). the only one with 730 mb porn on his quota (all in jpg), he called "test samples". *shrug* whenever you made a locate something | grep thisandthat - all you get was asian slu..... as bycatch. and his algorithm didn't work either (since it never occured in his vita). actually his name was dicken...
also i recall a theorem from computer visons lecture which was something like: "till now there is no computer vison system with power and capabilities of a biological vision system like the human - which complexity is too high to be totally understood even by todays knowledge" blabla retina + optical chiasma vs. neural networks
I'm not good at math, so I need you guys to help me with something. Based on the information above, I'm mostly interested in those parts of the internets where V < 10. Anybody know any sites where V < 30, or V < 30n, because those parts of the internets are probably even better, right?
Wait, the women in those guys' pornography have high heels growing out of their feet?
There's gotta be a fetish in there somewhere…
I wonder if this is anything like makeup or bandages, where it's only for caucasian nudity. If so, I see a steady increase in the sweet chocolatification of the internets.
Also, 246 gives me a boner.