Chinese Facial Recognition Payment System Has 99.8% Accuracy
Jokes are made over the Internet about the striking resemblance between Asians. I saw a photo having around 8 Asian people and the uploader asked few questions like “which of these are twins?,” “which one of these is sleeping?,” and “which of these are males/females?” Honestly, I couldn’t answer a single question because they all looked same.
If it is so hard for humans to differentiate between Asians, is it possible for a computer to tell the difference? And not only that, is it possible for it to accurately authenticate everyone?
The Chinese Academy of Science (CAS) started looking for answers to the questions mentioned above. As a result, they now have developed a framework that recognizes faces with 99.8 percent accuracy, as reported by Science and Technology Daily. The system is developed after a year-long research carried out by CAS.
The interesting feature of the facial recognition system is that it can even recognize and differentiate between twins (Asian twins too!). The users also do not have to worry about the changes in their facial features. For example, the system can identify correctly even after the face has gotten fatter or thinner.
The system uses its learning abilities to accurately figure out the changes that occur in the user’s face. So it might be able to identify you even after you have grown a beard.
The researchers have set up world’s biggest Asian face database that has more than 50 million records. Their system utilizes this database, in addition to the learning techniques, to achieve the near-perfect accuracy. The system is capable of analyzing a face from 91 different angles and can work in a dynamic environment.
Every framework services apps in achieving desired functionality. Hence, a further application of this facial recognition framework is also being developed by Zhou Xi, director of the CAS Chongqing Green Technology Research Institute and his team.
This app will let users make mobile and online payments by using their phone’s front camera and webcam respectively. Zhou said that once the users install this app on their phone or PC and link their bank accounts with the app, it will take one second to identify the payer.
To address security concerns, the system is able to detect that the user is not sending a held-up photo or video. Zhou confirmed that the algorithm used for facial recognition is reliable and secure.
The app is expected to be available in 2015.