AI detects pornstars in videos

AI detects pornstars in videos

Pornhub porn video site is a master in the art of being talked about. Even with regard to current technological trends oriented on the side of artificial intelligence and machine learning.

Pornhub claims more than 80 million visitors per day and more than 5 million videos. In order to improve and facilitate the search for videos, the site announces the launch of a " artificial intelligence model. "

Given the size of the site, it is not surprising that Pornhub relies on algorithms to first complete the work of the users themselves, who tag videos with tags.

A machine learning system was fed with official photos of porn stars and several thousand videos. He learns to recognize their faces. This learning is supervised by Pornhub users who validate or not the tags assigned on the videos. Based on this feedback from users, the system continues to learn and becomes better.


As part of a test phase throughout the past month, the system was able to analyze and label 50,000 videos. Detection is confined to the faces of 10,000 professional actors and actresses. This nevertheless raises concerns with the amateur / professional border which can be tenuous in this environment.

Over the next year, Pornhub says it wants to analyze its 5 million currently hosted videos, and extend identification beyond faces to automatically describe videos. Tags of the type "blondes," outside "and on the positions practiced are cited as an example in a press release.