second pic was 2 white couples fornicating:
"score": 92,
"resultcode": 1,
"porn": true
third pic - african lady with breasts exposed:
"score": 0,
"resultcode": 0,
"porn": false
Maybe it only works on light colored people.
I won't even bother with a picture where there may be some fellatio.
I used Face.com's API prior to it being shutdown to try and detect porn by measuring the % area taken up by a persons face. This sounded slightly feasible since a head shot should not be porn? But a close up of a BJ or a facial...
The only solutions for now is to manually moderate or use crowdsourced solution like
http://crowdflower.com/rtfm
Side note: There is a free 25 API calls per day option, but you have to enter a credit card. Suck
The guy is getting such hate on the comments on his site, yet his opening line
is that his choice was ill thought. Let him express his issues right?
I choose MongoDB for my last side project and while it was awesome working schema-less and developing the client facing part of the project was certainly quicker to deliver, i feel pretty lost on the analytics/BI side of it and couldn't say it better than him:
"Not having JOINs makes your data an intractable lump of mud"
I only wasted a few hours integrating face.com and used it for gender detection mainly. In the end a lot of african people showed as the incorrect gender so I'm okay with removing this option.
So his idea make sense :)