How Facial Recognition Works in Porn
It is hard to deny that the porn industry and technological advancements go hand and hand. Back in the mid-90s online pornography websites pushed the limits of what the internet was capable of handling. For those of you old enough to remember, the internet in the 90s was pretty dull. Websites were made in Notepad or MS Front-page, and they all sucked, everything was slow loading and “hi-resolution porn” was a picture 640 pixels wide, that is 0.3 Megapixel. Not 30 Megapixel, not even 3 Megapixel, but three-tenths of a single Megapixel and very few pictures were even that hi-res. As the technology that powered the internet improved, porn sites were able to provide the user with more and more to fap to.
The next great frontier in porn tech is Computer Vision, which powers facial recognition and augmented reality powered experiences.
Of course, USPornService uses facial recognition in our search engines allowing our users to search for adult models based upon a photo of their face
Identifying a Porn Star in Photos
1) Choosing a Photo
The first step, of course, is finding a pic with something sexy in it that we want to id. So let's start with this picture of two blonde lesbians enjoying some friendly licking together.
Now that we have our photo, the next step is to have the computer determine if there are any human faces within the photo. There are a ton of open source libraries like OpenCV
that are used by many. Of course, there are a ton of propriety ones too.
The face is cropped from the original picture with the eyes and mouth aligned in a common location for all photos. With these attributes in the same general area for all images, it makes the next steps easier for the computer.
Using the data points read in from the photo of each face, a unique digital 'faceprint' is generated for that specific face. A 'faceprint' is as unique as a human's fingerprint and can be reliably used to identify persons in photos or videos. Depending on what system you are using this faceprint is usually represented by a series of floating point numbers.
Now we can search the database of porn stars and return the names of each of the two blonde lez hotties. In our case we found:
You can watch the video from this shoot on PornHub
Building our Database
USPornService currently has over 18,322
models indexed in our facial recognition database. To acquire this information we had to build a sophisticated array of interdependent systems that collectively work together to keep our index current.
The first step is to write web spiders to crawl the internet and find porn; this is the same concept on how Google indexes the internet, by “surfing” the web and following links. Luckily we have extensive experience having written custom spiders for multiple Fortune 500 companies in the past.
As our spiders come across adult content we fetch the image, use our facial recognition system to determine if the photo is of a model we know about or not. If we do not know about this person we add it to a pending queue to research, if we do know about the model we add this photo/video/gallery to our known content and attempt to harvest useful demographic information such as the model’s name, date of birth, gender, nationality, etc. All of this information is added back to our ever-expanding database.
As of this moment, our system has spidered over 4,123,784
photos that contain a human face. This does not count the millions of photos that do not have a detectable face, and trust me that is a lot of porn!
We also have data-feed agreements with many porn producers, giving us access to photos of their models and demographic information.