This Chinese facial recognition start-up can identify a person in seconds

In China, facial recognition technology — biometric computer applications that automatically identify an individual from a database of digital images — is a part of daily life.

Already about 200 million surveillance cameras are scattered around the country — to track big spenders in luxury retail stores, catch identity thieves, prevent violent crime, find fugitives, catch sleeping students in the classroom and even snag jaywalkers. In fact, nearly every one of its 1.4 billion citizens is in China's facial recognition database.

AI companies believe surveillance and face recognition technology will make the country safer, and in the U.S. the tools are increasingly being used with law-enforcement agencies. But civil liberties advocates believe the issues of error and privacy may outweigh the security value.

Nevertheless, China has been plenty vocal about its plans to be the global leader in artificial intelligence by 2030, a market where the facial recognition piece alone is expected to garner $9.6 billion by 2022, according to Allied Market Research.

One of the companies making huge strides in this space is Shanghai-based YITU Technology, which has gained wide recognition for its Dragonfly Eye System, a facial scanning platform that can identify a person from a database of at least 2 billion people in a matter of seconds.

Ranked No. 20 on CNBC's 2019 Disruptor 50 list, YITU has raised more than $400 million from investors, such as China Industrial Asset Management, ICBC International Holdings and Sequoia Capital, and is currently valued at $2 billion. The security surveillance market is $120 billion in China alone, and the company now wants to export its product globally.

Co-founders Leo Zhu and Lin Chenxi launched the company in 2012. Zhu, who is also the CEO, received his Ph.D. in statistics from UCLA and was a student of Professor Allan Yulle, a disciple of Stephen William Hawking. He later did his postdoctoral research, concentrating on the study of brain science and computational photography, at MIT's Artificial Intelligence Lab.

Chenxi was a senior expert at Alibaba Cloud.

When Zhu and Chenxi launched YITU, the local police immediately began using YITU's Dragonfly Eye system to analyze surveillance videos and identify people and cars. By last year the company's technology was being used in more than 20 provincial public security bureaus in over 300 cities. Today its AI technologies are utilized in industries such as banking, retail, public safety and health care — last year it released what it claims is the world's first AI-based cancer-screening solution for lung and other cancers.

More from CNBC Disruptor 50:

US and China's race to lead the future of technology has come down to $6 billion
Inside Impossible Foods' mission to mass-produce the fake burger of the future
The biggest breakthrough in agriculture to help feed the planet may come from outer space

YITU is also expanding to help cities digitize data such as traffic patterns, energy supply information and infrastructure development. Now the company plans to move even further afield.

In January YITU opened its first international office in Singapore, where it plans to hire more than 50 researchers, and the company recently formed a strategic cooperation with local governments and various organizations in Britain in the fields of public security, finance and health care.

Although YITU's track record is outstanding thus far — it won first place in the 2017 Face Recognition Prize Challenge organized by Intelligence Advanced Research Projects Activity for its highly acclaimed facial recognition devices, which boast a 95.5% accuracy rate — the company still has work ahead to become the facial recognition leader in China.

Chinese start-ups Megvii Technology and SenseTime are considered to have the most powerful facial recognition systems in the world. Backed by Alibaba Group, Megvii is known for its open-source facial recognition platform, called Face++, which more than 300,000 developers are currently using to build their own face-detection programs. Valued at around $2 billion, Megvii recently raised $750 million in funds ahead of a planned initial public offering later this year, which could net as much as $1 billion. SenseTime, valued at $4.5 billion, just signed an agreement to build Malaysia's first AI park — a projected $1 billion project.

In January a study compiled by the UN World Intellectual Property Organization found that China and the U.S. both dominate the AI industry, with both countries leading in patents and academic research. Tech giant IBM has the largest AI patent portfolio, with 8,920 patents, ahead of Microsoft with 5,930.

Yet with respect to facial recognition, there have been more than 900 facial recognition patents filed in China — almost 10 times more than the number of patents filed in the U.S., according to data analysts CB Insights.

"I think that systems of governments make a huge difference here," says Kara Frederick, associate fellow for the Technology and National Security Program at the Center for a New American Security (CNAS). "[China] has really different corporate government practices than we do here in the United States."

The U.S. faces stringent privacy laws, in contrast to articles 7 and 14 of the National Intelligence Law of the People's Republic of China, which require intelligence cooperation between the Chinese citizens and their government.

Recent reports claiming that half of all U.S. adults are now on a facial recognition database has sparked notable controversy over privacy. Commercial companies have been slapped with more than 30 class-action lawsuits from consumers in Illinois alone in 2017 (Illinois was the first state to enact the Biometric Information Privacy Law in 2008, followed by Texas in 2009 and Washington in 2017). And in March, U.S. Senators Brian Schatz (D-Hawaii) and Roy Blunt (R-Missouri) introduced the Commercial Facial Recognition Privacy Act of 2019 — bipartisan legislation that would prohibit private companies from sharing consumers' biometric data without obtaining their consent.

In an even bolder move, San Francisco voted on Tuesday to make the city the first in the U.S. to ban police and other government agencies from using facial recognition technology, after supervisors voted 8 to 1 in favor of the "Stop Secret Surveillance Ordinance," which will require city agencies to disclose current inventories of surveillance technology.

Peter Trepp is CEO of California-based FaceFirst, a facial recognition software platform that works with law enforcement and across several industries, including transportation and retail, to prevent theft, violence and fraud. He believes that software providers have to be vigilant about security at all times.

In his book "The new rules of consumer privacy: Building loyalty with connected consumers in the age of face recognition and AI," he says, "We can't pretend that it is 1974, or even 2014. We have entered a new era that demands a new set of rules for both companies and customers."

Government agencies, including the FBI, have been using facial recognition technology to identify threats and prevent crime for more than a decade.

But now, through advances in artificial intelligence, facial recognition is expanding beyond law enforcement and into other sectors. Retailers across the country are integrating facial recognition technology into cameras to estimate a customer's age, gender or mood so stores can target them with ads on in-store video screens; banks are relying on facial recognition software to improve security and eliminate fraud. And researchers at Duke University developed an Autism & Beyond app that uses the iPhone's front camera and facial recognition algorithms to screen children for autism.

While Microsoft's facial recognition technology is still being deployed in many of the latest Smartphones to strengthen the security of devices, the company is advocating "for safeguards for people's democratic freedoms in law enforcement surveillance scenarios and will not deploy facial recognition technology in scenarios that we believe will put these freedoms at risk."

Trepp says FaceFirst is known for its performance "in the wild," which involves detecting the presence of known criminals in retail stores, stadiums, airports and other environments.

Among the data sets FaceFirst uses is the Labeled Face in the Wild, which is a database of face photographs designed for studying the problem of unconstrained face recognition, which closely resembles real-world situations. It is sponsored and managed by the Computer Vison Laboratory at the University of Massachusetts. The database is made up of 13,000 images, including variation in pose, lighting, expression, background, race, ethnicity, age, gender, clothing, hairstyles, camera quality, color saturation, and other parameters. Trepp said the latest accuracy score for this test was 99.97%.

He recalls "the fake beer deliveryman" as one of his most memorable cases: "A man dressed in the uniform of a well-known beer brand had developed a scheme where he would pretend to deliver beer to grocery stores, but in reality he was stealing it. ... Once the retailer's loss prevention team caught on to the scheme, they were able to use face recognition to detect him the next time he struck and work with local police to make an apprehension."

Yet Jake Laperruque, senior counsel at the Project on Government Oversight, isn't convinced the technology is accurate. "A lot of this technology is really really flawed," Laperruque says, especially for people of color and women.

And as private tech companies continue to store such large photo databases, Laperruque fears that it's only a matter of time before the government starts demanding access to those photos.

"Facial recognition is a privacy issue; it's a civil breach issue; it's also a civil rights issue," he says.

Correction: This story has been updated to reflect that Peter Trepp is the CEO of FaceFirst. Joseph Rosenkrantz founded the company in 2008 and held the position of chief executive until 2017.