Facebook May Recognize Faces Better Than You Do | Tier10lab
Image Image Image Image Image Image Image Image Image Image
Scroll to top


No Comments

Facebook May Recognize Faces Better Than You Do

Facebook May Recognize Faces Better Than You Do
Eric Huebner

Far from its humble beginnings as a way for Harvard students to keep in touch, Facebook has risen to dominate every facet of modern media. Not only does the social network boast over 1.23 billion active users, but it also controls other media juggernauts like Instagram, WhatsApp and, most recently, Oculus. Despite this clear industry dominance, Facebook isn’t taking time to sit back and relax. The company recently announced the development of a completely new type of facial recognition software. Although it’s only in its infancy, the new algorithm, entitled DeepFace, shows enormous promise, not just for easing the tagging of friends on Facebook, but for a host of other applications as well.

While the world’s largest social network already boasts a reasonably effective and predictive tagging function that suggests friends to tag when new pictures are uploaded, DeepFace represents a massive upgrade in terms of accuracy. Although there’s no definitive figure regarding the success rate of the current algorithm, DeepFace can theoretically identify a human face with 97.25% accuracy, roughly the same rate as a human.

This revolutionary software depends on a new 3D modeling technique developed by Facebook’s research team. To develop the technique, Facebook’s researchers analyzed 4.4 million tagged faces that were harvested from the uploaded photos of 4,030 unique users. The system was programmed to reproduce these faces as 3D models that can be rotated, viewed and analyzed from the varying angles that might present themselves in a photo.

Although there are certainly negative aspects to this software, foremost that it could make online anonymity vastly more difficult for those who desire it, it also represents an incredible opportunity for a variety of firms, particularly those involved in marketing and advertising. As digital marketing relies increasingly on user-generated data and profiles culled from cookies and search histories, facial-recognition software with this degree of accuracy could allow marketing firms to put together even more complete consumer profiles and allow them to engage at an even deeper level with consumers.  Despite this software only being in its infancy, its potential is sky high, not just for social networking but for the continuing development of the marketing industry.

[Sources: The Guardian, TechCrunch, Mashable]




4,342 total views, 2 views today