Why COVID-19 May Means Advancement in Facial Recognition Technology

New Innovative Facial Recognition Software Would Recognize User Even When Wearing Face Mask

0 8

Like most of the general population that started wearing face masks while being in public as a preventive measure for the coronavirus pandemic, numerous smartphone users faced trouble while unlocking their phones in public.

As technological advancement has come to a point where the user does not require to physically tap in a passcode to unlock their smartphone due to the fast and reliable facial recognition software being provided in almost every launched phone for the past few years.

Upgrading facial recognition algorithm

Many individuals like to unlock their smart devices by using facial recognition features while being in public, but the phones are unable to identify its user as the face mask is covering most of the lower half of their face.

Professor Hassan Ugail, who is an expert in the field of facial recognition at the University of Bradford, said that he faced an issue while unlocking his iPhone in public due to the donning of the face mask. He would rather just let himself unlock the device just by looking at his eyes.

Coincidentally, Prof Ugail took part in a research study published in a journal last year with his Ph.D. student, which suggests that the upper half of someone’s face is enough for specially trained software for facial recognition to work successfully on a human subject.

Those commercial unlock systems that were able to authenticate individuals by using facial recognition software that utilizes the entire face are now suffering due to the rise of the use of face masks amid the coronavirus pandemic.

The problem regarding the persisting issue of unable to unlock smart devices with facial recognition when wearing a face mask was highlighted in the monthly report for the month of July by the National Institute of Standards and Technology (NIST) in the United States.

The scientists were able to found errors in around 89 different software for facial recognition they tested when the lower have on the face, including the nose and mouth of the individual, were obscured, and in some of the cases from less than 1% up to 50%.

Research on various facial recognition software

Several top tech firms are currently working to further update their algorithms for facial recognition, so they are able to work while the user is wearing a face mask, especially in those situations where the facial recognition technology is used to confirm the identity of an individual, including providing access to secured areas and buildings.

According to an updated report from the National Institute of Science and Technology, they have been able to found multiple algorithms for facial recognition that have been reconfigured since the coronavirus pandemic began and are making fewer mistakes than before when analyzing faced with masks on. In some of the tests, the rate of error was about 10 times better than previous test results.

The computer scientist working at the NIST, Mei Ngan, said that the multiple tech developers are adapting their facial recognition algorithms to handle users with face masks. During the testing, Mei Ngan, along with her colleagues, used different varieties of shapes to represent the different variations of face masks available in the market, but not actual masks were used in this case.

This gradual progression of the innovation and advancement for a face mask is now gradually becoming more popular in the tech industry than before, as numerous companies are looking forward to installing contact-free interaction with security devices that make it more hygienic in public places.

The personal software for facial recognition developed by Prof Ugail is basically used for research purposes, which works on the face area of the upper half and around the eyes. The results are yet to be published, by the accuracy rate of the software is around 86%.

Read also Elon Musk’s Starship Prototype Makes A Huge Impact

Leave A Reply

Your email address will not be published.