Tech & Gadgets

Facebook?s DeepFace Facial Verification Software Is Creepily Accurate

By on


Ever noticed that tagging your friends in your photos becomes simpler and more accurate in Facebook?

Facebook?s newest research on facial accuracy is attempting to copy the human ability to associate and match faces with 97.25 percent accuracy. Human accuracy is measured at 97.5 percent.

Technology Review earlier reported that the social networking giant has developed its innovative facial verification software that is capable of matching the correct persons with their pictures, regardless if the person faces the camera. This is done with a 3D map of the person?s face. The DeepFace algorithm knows how to recognize patterns from huge amounts of data in the Facebook database.

This novel artificial-intelligence technology, dubbed as ?deep learning,? enables Facebook to accurately match people with their photos, even if the subjects of the photo re not facing the camera, the expression is varied, or the lighting is not ideal.

Yaniv Taigman of Facebook?s AI team says that this sort of development that closely approaches human performances and capabilities is not normally seen. DeepFace?s software for facial verification matches two images wherein the same face shows, instead on recognizing a person based solely on facial characteristics.

Facebook?s facial recognition algorithms are used to suggest which of one?s friends should be tagged in photos. The tag suggestion, first seen in June 2011, has since met with privacy advocates worldwide.

Privacy Issues

This software means that one?s photos, taken by another, can be instantaneously tagged to their profiles. Naturally, various privacy issues have stemmed, especially since there are claims that many governments already has a hand on each individual?s account.

Facebook states that to develop the software, it used a pool of over 4.4 million faces labeled by over 4,000 of its users. While the software is not yet used in Facebook per se, its presentation to the developer community is meant to garner various feedbacks from other programmers and researchers.

Back in 2012, the European Union made Facebook delete all its facial templates collected from the area. Currently, Facebook?s facial recognition is unavailable for users in the United Kingdom.

About the author

To Top