Tagging friends and family on a Facebook photo update could soon get a lot more accurate with Facebook currently on the look out to ace in domain of facial recognition. Achieving near human capability of recognizing face was hitherto considered beyond the current technological advancement, however, DeepFace is set to change all that under the watchful eyes of its developers.
A recently published paper by Facebook’s R&D explicitly detailed the internet giant’s exploits in this arena. Christened as DeepFace, researchers have claimed that this technology has the capability of identifying faces accurately up to 97.25% of the time it was put to test. For this, two mutually exclusive photographs were scanned by DeepFace before it delivered the results to the researches. In comparison, human capability is as high as an estimated 97.53% on an average and even higher for people with an eidetic memory.
DeepFace’s accuracy is derived from its ability to create a three-dimensional model of any human facial structure and then scanning Facebook’s network to complete the identification process. This generation of 3D model created from an available photograph has been designed by the programmers at Facebook and aptly named DeepLearning.
Facebook, although, is currently using an advanced facial recognition technology which was also developed by its researchers. The same can be seen in action regularly by the patrons of the social networking website when it suggests names from the user’s network for tagging immediately after an upload is initiated. However, DeepLearning’s algorithm to identify facial features is being touted as far more advanced than what is being used by any company at this point in time, which also includes Google’s G+.
According to the researches, what helps DeepFace in acing the technology is the fact that while other available software are incapable of scanning a face in a photograph unless it gets a clean frontal view. However, with three-dimensional rendering of the photos, DeepFace is able to create a more accurate model for it to compare and look for possible matches in its database.
Indulging a little more deeply in to the technology being used, the researches detailed the current system of facial recognition into four critical steps. The software first detects a face, then aligning it with the data base, representing the data and finally classifying it based on the stored information. For DeepFace, developers focused greater attention to the stages of aligning and representation through the 3D modelling structure with multi-layered networking structure.
With over 120 million parameters to be cross-referenced during analysis of the generated 3D model, DeepFace showcases a massive jump in the segment of facial recognition system. Although, this technology could have far reaching benefits for Facebook, the spokesperson clarified that the company does not intend to induct DeepFace into the website’s armory anytime in the near future.
This particular addition to the announcement could be a result of the massive recent outcry against privacy concerns raised by worried patrons. Facebook was under fire earlier this year after being accused of colluding with snooping government agencies, which in the current scenario, could result in a lot of raised eyebrows in the wake of DeepFace’s capability revelation.
However, with the researches at Facebook claiming to have bridged the gap between human and machine capability to recognize faces, time has come to further indulge in the possible implications of such technology. It is now required to undertake a deep study on the subject and if required possible regulations put in place to look out for possible breach in privacy, especially by addressing the concerns of the patrons. Till then everyone can only hope that no unwanted information leaks are initiated with DeepFace acing its way into facial-recognition-software history.
By Daris Abraham