PornHub uses computer vision to ID actors, acts in its videos

They say porn adopts new technologies first and PornHub [NSFW], a site that is as NSFW as it sounds, is proving that old adage true. The site – which receives 80 million visitors a day – found that its old, antiquated methods of tagging videos by hand was not sufficient.

Rather than hire someone to go through the hundreds of thousands of videos and tag them by hand, the team at PHub hired a robot.

“Ultimately, we want to provide our fans with whatever it is they have a penchant for and our new model will be able to surface more accurate results for them, which will hopefully keep them coming back for more,” said Corey Price, PHub VP. “It’s all about continuously updating our platform to provide our fans with the latest in technology to make their experience on our platform as seamless as possible. The easier it is to navigate and access what they want, the happier they will be.”

The computer vision system can identify specific actors in scenes and even identifies various positions and… attributes. While it is obviously very difficult to describe the feature set for a family audience, the system can identify individual performers in real time – in the demo here it recognizes one performer even from the side – and it can also identify sex acts.

“So far we’ve used the model on about 500k featured videos which includes user submitted and we plan to scan the whole library in the beginning of 2018,” said Price. “Very shortly, the technology will also be used to detect various sex positions / categories and be able to properly tag them as well.”

Facial detection is nothing new, even for mobile devices, but this system goes one step further by categorizing videos and images based on various attributes. This means you’ll be able find favorites by name or characteristics, a feat that once require prodigious amounts of data entry. How well does it work?

“It’s quite precise,” said Price.