After thousands of man-hours invested in the feature of facial recognizing, both Google and Facebook have gotten really good at it. Their algorithms have become so advanced that your face doesn’t even have to show up in the picture; the back of your head will soon be enough for Facebook to know it’s you.
When it comes to facial-recognition algorithms, Facebook has become a leading organization, with such innovative technology that it can now identify people in pictures as well as humans can.
With an astonishing 97 percent accuracy, Facebook’s algorithm dubbed DeepFace (really, Facebook?) can tell if it’s you in two different pictures, without any other source than the pictures themselves. Not even FBI’s identification system has such good scoring in accuracy.
DeepFace has received its training on one of the largest facial databases which incorporates more than 4 million facial pictures of roughly 4,000 individuals. Therefore, it’s gotten really good at analyzing faces by turning them into 3D models: this step is important, because it allows the algorithm to recognize the same face from different angles and under various lighting conditions.
According to an article on Facebook’s research, more than 120 million parameters are involved in this kind of facial recognition algorithms. But Facebook wants to take things to the next level: identifying people even when their faces are visible. In order to do this, the algorithm is trained to focus on body shape, hairstyle, clothing and posture.
With this upgraded algorithm, Facebook reported that the accuracy has reached 83 percent in cases of pictures of people whose faces aren’t showing. So the public wouldn’t freak out hearing the news, the giant tech company researched and developed the tool by using Flickr shots and not actual Facebook photos.
While Facebook’s algorithm is definitely impressive, Google has not been idle either. A new feature for Google Photos was announced last month and it has left people with their mouths wide open.
It is absolutely crazy what Google Photos’ search engine can do; not only can it differentiate between pictures of cats and dogs, but also search based on dog breeds. And this is the least of its capabilities.
Photos can also be searched and categorized based on a key word or an adjective that can describe the images you’re looking for. A demo showed that typing in “delicious” in the search bar promptly returns results of pictures of food and beverages.
Recognizing cats, dogs and food is all fun and games, but even when it comes to people Google Photos delivers. The search engine makes it clear that your searches can include “People, Places and Things,” and when you click the More link on the People option, you’ll have a list of all the people you’ve ever taken a photo of.
Such news advancing the recognition capabilities are becoming increasingly worrying – even for people who think nothing of the NSA’s blatant privacy invasion. There are plenty of applications for Facebook’s and Google’s algorithms outside identifying people in pictures, and the government would certainly be delighted to get its hands on either of them.
Image Source: IT ProPortal