Facebook and Facial Recognition Software....

Michael Bilson
Found this article interesting and curious what your thoughts are on this topic. As far as facial recognition technology goes, you’d be forgiven for being surprised that Facebook is making serious inroads in the industry. After all, this is a social network we’re talking about - not MI5 or the FBI. However, the somewhat ominously named ‘DeepFace’ project being rolled out by Zuckerberg's company is doing just that, with startling accuracy. Facebook claims that 400 million photos are added to their site every single day, they’re now able to tell when you appear in any of these photos. The extraordinary accuracy of the software is not only raising eyebrows around the world, it's also raising privacy concerns. But how does all this work, and what will it mean for you, humble Facebook user? You may have noticed that Facebook already has rudimentary facial recognition technology in operation, upload a photo and it occasionally will recommend faces that it finds familiar to tag. They’ve now taken some bold steps forward and their algorithms are now ridiculously accurate - they can detect faces with 97.25% accuracy no matter the condition of lighting or angles. To put this in perspective, humans have an average of 97.53% accuracy when identifying faces. What Facebook has done here is design software that is on a par with the human consciousness. This technology is partly the brainchild of Yaniv Taigman, one of Facebook’s artificial intelligence scientists who co founded Face.com in 2007. The site was since acquired by Facebook who implemented their technology into their own rudimentary facial recognition software. The advancements are thanks to an area of AI development known as ‘Deep Learning’. Things get a bit heavy here but basically ‘Deep Learning’ software mimics the neuron activity in the neocortex of the brain where ‘thinking’ as we know it takes place. The software learns to spot patterns in the digital embodiment of sounds and images and other stimuli - in a sense mimicking the way humans learn. The complexities of ‘Deep Learning’ is an issue for a different time, what’s relevant now is how it’s being implemented in a commercial setting, namely by Facebook in this instance. DeepFace processes images by first correcting the angle of a face so that it appears the person is face-on towards the camera using a standard 3D model of a straight-on face. The ‘Deep Learning’ aspect of the technology then comes into play by using its simulated neural network to assign numerical values to aspects of the face. The final step is to compare that to the original image to determine the similarity. In a sense it’s ‘Facial Verification’ rather than ‘Facial Recognition’ at this stage as it can only compare two images, not match a face to a name. Facebook have stressed that DeepFace is purely for research purposes at this stage, although the potential in this technology is quite astounding. For you, this could mean fundamental changes in security features for your technology in the future - imagine never having to use keys to get into your home, just standing by the door and letting the technology do its stuff. For the industry this means gaps appearing for software engineers, AI scientists and cyber security professionals - as well as host of other areas within the field. Science fictions writers dreamed of the day when computers could surpass human senses, are we seeing the dawn of the AI era? Link to article. https://www.linkedin.com/pulse/facebooks-facial-recognition-accurate-human-david-di-domenico
mark rask
wow this is real big brotherish!

Featured Masters

No members found

Other Marketing Solutions Products

100%  Recommended Recom'd 2 Ratings
Company: Outsell
100%  Recommended Recom'd 1 Ratings
Company: DealerSuccess
100%  Recommended Recom'd 1 Ratings
Company: DealerSuccess
100%  Recommended Recom'd 1 Ratings
Company: MAXDigital

 Rate a Vendor Give feedback in three quick steps

Select a Vendor & Product
Can’t find what you're looking for?
Add a Vendor  or  Add a Product
  •  
  •  
  •