Masks and facial recognition will be common in the future. How will they coexist?
Author: huifan Time: 2020-12-09
Surprisingly, public opinion can change quickly. As we count down twelve months, many of us will look suspiciously at a masked man.
Now, some countries have stipulated the use of masks in the law. Victoria has also imposed mandatory regulations on them, and they are also recommended in several other states.
One of the results is that facial recognition systems used for security and crime prevention may no longer be able to achieve their goals. In Australia, most agents remain silent on the use of facial recognition.
But documents leaked earlier this year showed that the Australian Federal Police and state police in Queensland, Victoria and South Australia all use the commercial facial recognition platform Clearview AI. The New South Wales Police also admitted to using a biometric tool called PhotoTrac.
What is facial recognition?
Facial recognition involves using calculations to recognize faces in images or videos and then measuring specific facial features. This can include the distance between the eyes and the relative position of the nose, chin, and mouth.
This information is combined to create a facial signature or contour. When used for personal identification (such as unlocking a mobile phone), the image in the camera is compared with the recorded profile. The process of facial "verification" is relatively simple.
However, when using facial recognition to recognize faces in a crowd, it requires an important profile database to compare with the main image.
By registering a large number of users into the system, these configuration files can be collected legally. But they are sometimes collected secretly.
Facial "verification" (the method used to unlock a smartphone) compares the main image with a pre-saved facial feature. Facial "recognition" requires checking images against the entire facial feature database. Teguha Tiplas (pixguy)
The problem with masks
Since facial features are based on a mathematical model of the relative positions of facial features, anything that reduces the visibility of key features (such as nose, mouth, and chin) will interfere with facial recognition.
There are many ways to evade or interfere with facial recognition technology. Some of them are developed from technologies designed to circumvent license plate recognition systems.
Although the coronavirus pandemic has heightened concerns about evading facial recognition systems, leaked US documents show that these discussions also took place in 2018 and 2019.
Although the debate about the use and legality of facial recognition continues, the recent focus has shifted to the challenge of wearing masks in public.
In this regard, the National Institute of Standards and Technology (NIST) coordinated a major research project to evaluate how masks affect the performance of various facial recognition systems used worldwide.
Its report released in July found that certain algorithms had difficulty correctly identifying people wearing masks up to 50% of the time. This is a large error rate compared to the same algorithm when analyzing uncovered faces.
When the mask covers too many faces, some algorithms are even difficult to locate.
Find a solution to the problem
There is currently no available photo dataset of masked people that can be used to train and evaluate facial recognition systems.
NIST's research solves this problem by overlaying masks (various colors, sizes, and positions) on facial images, as shown below:
Although this may not be a true portrayal of people wearing masks, it is sufficient to effectively study the impact of masks on facial recognition systems.
An image of a real masked person may allow more details to be extracted to improve the recognition system-perhaps based on the protrusions visible in the mask to estimate the position of the nose.
Many facial recognition technology providers have prepared for the future where the use of masks will continue or even increase. An American company provides masks with the faces of customers printed on them so they can unlock their smartphones without having to remove them.
Motivation to wear masks continues to increase
Even before the coronavirus pandemic, masks were a common defense against air pollution and viral infections in countries including China and Japan.
Political activists also wear masks to avoid detection in the street. Both the Hong Kong protests and the Black Lives Matter protests have increased the protesters’ desire to evade the authorities and government agencies for facial recognition.
As experts predict, there will be more epidemics in the future, air pollution levels will rise, the persistence of authoritarian regimes and the increase in bushfires that are expected to produce dangerous smog – wearing masks may become the norm for at least some of us.
Such technology is already under development. Multiple vendors are providing upgrades and solutions, claiming that reliable results can be obtained in subjects wearing masks.
For those who oppose the use of facial recognition and want not to be discovered, an ordinary mask is enough now. But in the future, they may have to consider other options, such as masks printed with fake computer-generated faces.
If you want to read more interesting articles about artificial intelligence, its applications and how it changes our world, please stay tuned to hfteco.com
Huifan Technology successfully customized a variety of requirements for facial recognition temperature for customers