How China built facial recognition for people wearing masks

When it’s come to the most advance facial recognization system no one can beat China, the country has 170 million closed-circuit television (CCTV) cameras, with another 400 million expected in the next three years.

Most of these cameras are equipped with AI and facial recognition. At the same time, in 2017, China filed for more than 900 facial recognition patents, compared with the 96 patents filed in the U.S.



Right now Chian is heavily suffering from the damage caused by coronavirus, the virus has 3,248 death in China. Even in this difficult time, China has taken its facial recognization system to a new level. A Chinese company has created a new facial recognition system that can identify people even if they are wearing masks.

The company that builds this new facial recognition system is named Hanwang. Hanwang manufactures and produces pattern recognition technology, the company has placed 2 million of its cameras across the world.




Hanwang has started working on the new system in January when cases of the new coronavirus began rising in China. The fast spread of the virus has led most Chinese citizens to wear face masks in public.

Hanwang is a Beijing-based firm. Huang Lei, the company’s chief technical officer, said that even before the new virus was widely known, he had begun to get requests from hospitals at the center of the outbreak in Hubei province to update its software to recognize nurses wearing masks.




Huang told Reuters: If connected to a temperature sensor, it can measure body temperature while identifying the person’s name, & then it would process the result, say if it detects a temperature over 38 degrees.

Engineers at the Beijing-based Hanwang Technology Ltd. say their system is the first to be created to effectively identify people wearing face masks. Hanwang Technology used a sample database of around 6 million unmasked faces and a much smaller database of masked faces to create the system.

The company said that a team of 20 people built the system in a time period of a month. The facial recognization system is based on existing technologies developed over the past 10 years.

The company says its masked system has reached 95 % accuracy in lab tests & claims that it is more accurate in real life, where its cameras take multiple photos of a person if the first attempt to identify them fails.

“We wouldn’t wait until something explodes to act. If three or five clients ask for the same thing . . . we’ll see that as important,” said Mr. Huang, adding that its cameras previously only recognized people in masks half the time, compared with 99.5 percent accuracy for a full-face image.

Hanwang is now selling two main kinds of products that use new technology. One performs “single-channel” recognition, which is designed to be used at the entrances to buildings.

The other product is a “multi-channel” recognition system that uses groups of surveillance cameras. Huang Lei said that the multi-channel system can identify individuals in a crowd of up to 30 people “within a second.”

“The problem of masked facial recognition is not new, but belongs to the family of facial recognition with occlusion,” Mr. Huang said, adding that his company had first faces similar issues with people with beards in Turkey & Pakistan, also with northern Chinese customers wearing winter clothing that covered their face.



While training facial recognition algorithms to recognize masked faces involves throwing data away. A team at the University of Bradford published a study last year showing they could train a facial recognition program to accurately recognize half-faces by deleting parts of the photos they used to train the software.

When a facial recognition program tries to recognize a person, it takes a photo of the person to be identified, & reduces it down to a bundle, or vector, of numbers that describe the relative positions of face.

But as the algorithm is applied against a larger population, there is a greater chance of misidentifying a masked face, since there is less information to work with compared with full faces and there may be multiple people with similar features around their eyes and nose.

In Hanwang’s case, Mr. Huang said that the company’s devices were designed to work in office settings with a database of up to 50,000 employees’ faces. He said the system was able to use photos taken from the Chinese police’s identification card database of some 1.2 billion people.

When a facial recognition system has calculated its vector of facial features for the person to be identified, it compares it to the vectors of the faces in its database, finding a match if it meets a certain degree of accuracy.

Hanwang’s system works for masked faces by trying to guess what all the faces in its existing database of photographs would look like if they were masked.

To improve its guesses, Hanwang asked its 2,000 or so employees for photos of themselves wearing masks, but also created another database of 6m photos of people with artificially generated masks — with different styles to reflect the ones commonly worn in China.

The company says one of the largest users of the new system is China’s Ministry of Public Security, which operates the country’s police agencies. The Chinese government widely uses cameras and facial recognition technology for identification purposes and to record people’s movements.

Huang said officials can use Hanwang’s technology to compare images with ministry records on individuals in order to identify and track people as they move about. The system can identify crime suspects, terrorists or make reports or warnings, he added.

However, the new system struggles to identify people wearing both a mask and sunglasses. “In this situation, all of the key facial information is lost. In such cases recognition is tough,” Huang said.

More in AI

Researchers Use AI to Create Music through Proteins

Google developing a tool to help triage individuals for Covid-19 testing

Google Ph.D. fellowship programme: Indian students can apply by April 15

Microsoft’s New AI Generates 3D Objects From 2D Images


 

Leave a Reply

Your email address will not be published. Required fields are marked *