DeepFace is Facebook’s new weapon in the making for recognizing faces in the images uploaded there, and it is now very nearly as accurate in detecting facial features as our brain. The software looks at two images, and irrespective of lighting effects or angle, it can with more than ninety percent accuracy identify the same face. DeepFace happens to be in its development days, but in the future it will likely be used to for facial recognition on Facebook.
The DeepFace program, created by social media giants artificial intelligence research group that is based in California, is underpinned by a sophisticated deep learning neurological network. A neurological network is a piece of software that simulates a practical approximation of how a real neuron system or grid functions. Deep learning is among many methods of performing machine learning; primarily, it goes over an extensive body of data (for example, people’s faces) and tries to come up with a high-level generalization of an individual’s face by searching for recurring patterns like eyes, ears, lips, etc.
As soon as the learning process is complete, every picture that’s fed into the software passes through the grid of the neural network in a different way, creating a distinctive fingerprint. For example, a single neuron could ask if a person in the picture has blue or brown eyes — if so, route x is followed, if no, another course is chosen.
Furthermore, the AI software when comparing two pictures belonging to the same person can verify a match with an accuracy of more than ninety percent. Men and women, performing the same verification experiment with a similar group of images, scored relatively better. However, DeepFace isn’t impacted by varied light from the two photos, as well as pictures taken from odd angles are instantly transformed by the software resulting in a three-dimensional with the face looking forward – to ensure that all comparisons are done with a standardized, forward-looking picture. The data indicates that efficiency — vital factors when talking about the usefulness of machinery based algorithm — looks excellent.
However what Facebook chooses to do with the system, needless to say, remains a secret. It might apparently be used to shore up the existing facial identification options, making sure that each picture of a person is connected to their respective account(s) even if they are not tagged. From there, it is hard to imagine that Facebook will keep DeepFace strictly limited to Facebook — there is too much money to be earnt by scanning the remainder of the open internet for matches. Additional it might be used by the government agencies and marketing firms for studying shopping behavior, let’s just hope for the best
Source: Business Insider, The Hacker News
You want to support Anonymous Independent & Investigative News? Please, follow us on Twitter: Follow @AnonymousNewsHQ
This Article (An Idea That Can Change The World) is free and open source. You have permission to republish this article under a Creative Commons license with attribution to the author and AnonHQ.com.
