Google Says Facial Recognition System 'Nearly 100 Percent Accurate'

A new facial recognition system from Google called FaceNet can pick a face out of a crowd — in fact, it can pick a face out of a crowd of 260 million.

And it could mean that the type of software seen on hit TV shows like "NCIS" and "The Blacklist" will be widely available in the near future.

The surveillance-enhancing system could be used to catch criminals or track terrorists as well by dating sites or social media marketing analysts, according to Fortune.

Last week, Google researchers released a paper saying that the artificial intelligence system dubbed FaceNet represents the most accurate approach yet to recognizing human faces.

Scanning a massive facial-recognition data-set called "Labeled Face in the Wild," FaceNet was able to identify faces from across the Web out of 260 million images "with nearly 100 percent accuracy," the report stated.

The researchers were testing for how impressive the algorithms are at deciding whether two images are of the same person, Fortune reported. The study also tried to verify whether FaceNet can put a name to a face, as well as provide collections of faces that look the most similar or the most distinct.

In the case of dating sites, for example, a woman might be looking for a George Clooney lookalike from around 2000, and the system might come up with dozens of profiles in the area.

Last year Facebook researchers published a paper saying that it has a 97 percent accuracy rate with its DeepFace face recognition system.

Fortune says that FaceNet and DeepFace, both still in the research stage, make it easier for users to tag photos and search for people. They will also allow Web companies to analyze their users' social networks and to assess global trends and celebrity popularity.

Related Stories:

• Facial Recognition Nabs Army Fugitive After 37 Years on Lam
  
• Anne Boleyn Portrait Found? Facial Recognition Run on Painting