Internet Giants Are Perfecting Facial Recognition Algorithms — Google’s FaceNet Can Recognize You With 99.96 Percent Accuracy

Internet giants are working on perfecting their own facial recognition algorithms and they are getting scarily accurate. Google’s own FaceNet can identify your face with an astounding accuracy of 99.96 percent.

Soon, it will be nearly impossible for anyone to prevent computers from identifying you, even if you are hidden in a barrage of human faces. While Facebook’s facial recognition algorithm, which the company fondly refers to as DeepFace, currently manages to identify you with 97.35 percent accuracy, Google’s own offering, called FaceNet, seemingly will never make a mistake as its algorithm can identify a face with 99.96 percent.

Soon You Won't Be Able To 'Disappear' Into The Crowd

Needless to say, Google’s FaceNet is nearly or more importantly, practically perfect. No wonder, Google researchers call their system the most-accurate technology available for recognizing human faces.

Though Google has admitted to a miniscule possibility of error, in its paper, titled “FaceNet: A Unified Embedding for Face Recognition and Clustering,” Google claims the system achieved nearly 100 percent accuracy rate on the facial recognition dataset Labeled Faces in the Wild. The database comprises of 13,000.jpeg images of human faces and Google’s FaceNet managed to identify the right person almost all the time. Interestingly, when FaceNet was tested against YouTube’s dataset, it was only 95.12 percent accurate.

Google's FaceNet Is The Most Accurate Facial Recognition Software Today

What’s concerning is the fact that the algorithm isn’t affected (or can be cheated), by altering facial features. In their current iterations, Facebook’s DeepFace and Google’s FaceNet, seem to have minor issues in accurately accounting for different poses and light levels.

The next set of tweaks will use the minuscule error rate to further improve the accuracy, explained one of the researchers,

“Our method uses a deep convolutional network trained to directly optimize the embedding itself, rather than an intermediate bottleneck layer as in previous deep learning approaches. To train, we use triplets of roughly aligned matching / non-matching face patches generated using a novel online triplet mining method. The benefit of our approach is much greater representational efficiency: we achieve state-of-the-art face recognition performance using only 128-bytes per face.”

Apparently, FaceNet was trained on an enormous 260-million-image dataset and performed at an 86 percent and higher accuracy rate. Meanwhile, a team of Chinese researchers claimed they had managed to achieve better than 99 percent accuracy.

Light Conditions And Poses Are Perhaps The Only Challenges That FaceNet And DeepFace Face

Google’s FaceNet is more potent than Facebook’s DeepFace for a very prominent reason. FaceNet not just verified if the two faces were identical, it also managed to “identify” the person in the image. While FaceNet has been developed by Google, DeepFace uses technology designed by an Israeli startup called face.com.

Apart from these two internet giants, other companies like Yahoo and Microsoft are said to be deep into facial recognition as well. Worryingly, the makers claim the technology is merely meant to “tag” people in photos making “discovery of content easy”.

[Image Credit | Google, Facebook, Shutterstock, Galleryhip, Fanpage]