Coincidentally I was trying this out yesterday as a potential option for automated porn filtering for the imageboard that I administrate. It is very lightweight which is good but the performance is not great. Some weird false positives and it does a bad job with flat chested women which is very concerning to me because it leads me to think it won’t be effective against CSAM.
unglaublich 26 minutes ago [-]
I guess CSAM detection will remain hard as the training data raises so many ethical concerns.
I wonder if there's a way in which we can forward-hash sensitive material like CSAM and train a classifier that only consumes the 'hashed' version of the material for training and detection.
tao_oat 1 minutes ago [-]
I believe this is how many CSAM classifiers work. There isn't a simple legal avenue to host CSAM for classifier training, for good reason.
RobotToaster 1 hours ago [-]
Weird, with it being from the UAE I'd expect it to detect any woman with an exposed ankle as porn.
gumby 47 minutes ago [-]
Have you been to the UAE? Sure t the beach there re women in burkinis and more, but also women in two piece bikinis.
wolvoleo 45 minutes ago [-]
There's a big difference between international jersetters partying in Dubai and locals.
The jersetters are basically exempted from the strict laws because they bring in money.
Alcohol is illegal there, homosexuality is, even unmarried sex is.
I certainly will never visit there even if they choose not to apply those laws to me as a visitor.
figmert 45 minutes ago [-]
What a strange Islamophobic comment to make. You should probably go out a little more.
echelon_musk 11 minutes ago [-]
Afghanistan.
instagraham 2 hours ago [-]
it is amusing that a model from a country that has banned porn would ever be developed. Wouldn't there be hurdles for them to access "data" at every stage?
(I know they must have used a VPN for training but it does seem like an added hoop. OFC the UAE's block is also not particularly thorough.)
You'd assume such models would do better in a laissez-faire nsfw environment than in a walled garden of relatively-less sin.
dfish 1 hours ago [-]
It's for research purposes :wink: :wink:
waffleiron 32 minutes ago [-]
There are plenty western CSAM detection models
DC-3 20 minutes ago [-]
Can you suggest a good one?
sillyfluke 1 hours ago [-]
It's a story as old as time so I'm frankly less amused by it. The old school censors always saw every "data" that was taboo under the sun prior to censoring it for public consumption.
ggm 3 hours ago [-]
Always two sides to classification. Finding and Banning. It's tempting to ask if this is one, the other or both by the same people.
totetsu 2 hours ago [-]
Planet-scale nude-selfie extenuation racket?
londons_explore 46 minutes ago [-]
How many people download classifiers?
The number of downloads really doesn't seem like a good way to measure the quality of a bit of work. One download could be deployed on a bit website and classify billions of users images.
londons_explore 45 minutes ago [-]
Smells like someone is botting the download numbers to get a press release....
nlanandkumar 2 hours ago [-]
[dead]
Rendered at 09:47:34 GMT+0000 (Coordinated Universal Time) with Vercel.
I wonder if there's a way in which we can forward-hash sensitive material like CSAM and train a classifier that only consumes the 'hashed' version of the material for training and detection.
The jersetters are basically exempted from the strict laws because they bring in money.
Alcohol is illegal there, homosexuality is, even unmarried sex is.
I certainly will never visit there even if they choose not to apply those laws to me as a visitor.
(I know they must have used a VPN for training but it does seem like an added hoop. OFC the UAE's block is also not particularly thorough.)
You'd assume such models would do better in a laissez-faire nsfw environment than in a walled garden of relatively-less sin.
The number of downloads really doesn't seem like a good way to measure the quality of a bit of work. One download could be deployed on a bit website and classify billions of users images.