Powered by artificial intelligence and thousands of cameras, a Ukrainian computer vision startup is creating a future where every person can be recognized on the street.
Called Traces.AI and developed by Ukraine-born Veronica Yurchuk and Kostyantyn Shysh, its technology analyzes over 2,000 of a person’s physical traits through security cameras – from their hairstyle and the accessories they wear to the texture of their clothes.
The startup’s surveillance technology, which uses videos from the CCTV cameras of its clients, has successfully been used to find missing people in seconds, despite having only a short description of their clothing or appearance.
And with new challenges on the horizon, Yurchuk and Shysh have gone even further: They have developed a tool to track people who came in contact with COVID-19 patients and could themselves be carriers of the novel coronavirus.
The creators of this technology claim it could curb the mass spread of COVID-19 if massively deployed right now.
Novel solution for novel virus
The idea to use security videos to fight the virus didn’t come to the founders by chance.
In the battle against COVID-19, many countries are trying to implement a “test and trace” model, successfully adopted in South Korea, Taiwan and Singapore. Governments monitor the self-isolation of people who have the virus via mobile apps and Bluetooth. Some send text mesages to people who have been in the vicinity of an individual diagnosed with COVID-19, asking them to self-isolate too. This approach allows the authorities to relax the quarantine restrictions, while still controlling the spread of the virus.
Other countries, however, still rely on a more traditional approach: Authorities just interview COVID-19 patients about people who could possibly come in contact with them. But this model is labor-intensive and inaccurate, Shysh said.
To offer an alternative, Shysh and his partner came up with a more reliable way of tracing COVID-19 — by analyzing video from security cameras. The creators claim the technology is 98% accurate and works effectively regardless of weather conditions and the quality of the cameras.
Traces.AI can also analyze data from thermal imaging cameras that are sensitive enough to detect a heightened temperature in humans, the founders said. These types of cameras are usually installed in airports and security organizations like the U.S. Defense Department. U.S. tech giant Amazon also started to use thermal cameras on its premises to detect people with temperatures who could potentially have COVID-19.
In both cases – with ordinary security cameras and thermal ones – their clients upload videos to the system, the system blurs the faces and sends the videos to the cloud. Then, the startup’s algorithms analyze them and detect certain features of a person, creating a digital model of the individual. The model is so accurate, it can locate and identify people within 2-3 seconds.
Although the technology gives almost immediate results, it takes an hour to verify all the contacts of the infected person. Nevertheless, the startup can identify the contacts of at least four different infected people simultaneously.
According to the founders, the accuracy of the analysis depends on the number of cameras installed indoors or outdoors. For a metro station, Shysh said, 16 cameras are enough to correctly identify the contacts, while Traces.AI can work with the video feeds of over 20,000 cameras at once.
At two factories in New York and Texas, the company’s clients, some employees tested positive for COVID-19 and the Traces.AI algorithms identified everyone who interacted with them fast enough to send them into self-isolation before more people got infected. Shysh wouldn’t disclose the names of the businesses because the startup works with them under non-disclosure agreements.
Privacy concerns
Experts expect that, by 2021, over one billion surveillance cameras will be installed around the world. As of 2019, more than 770 million cameras were already in place, half of them in China.
The adoption of facial recognition, one of the most popular surveillance tools, raises many privacy concerns and has already been banned in the U.S. states of Oregon and New Hampshire, the cities San Francisco and Oakland in California and the city of Somerville in Massachusetts.
The idea of privacy protection was central to Traces.AI, its founders claim. When compared to facial recognition technology that is subject to racial and gender biases and can be used to read human emotions, Yurchuk and Shysh believe that their algorithms identify people regardless of their ethnicity and sex. Their approach is less invasive to the person’s privacy than tracing mobile apps, which, apart from video images, receive people’s geolocation and potentially other personal data that is stored in their phones.
Before sending a video to the cloud, Traces.AI blurs out people’s faces, evaluating other physical features instead. “It creates a person’s unique profile that can later be tracked by different cameras in different parts of a city,” Shysh told the Kyiv Post.
The absence or replacement of one of the parameters doesn’t impact the general results because there are hundreds of other features it evaluates.
Promising field
Yurchuk and Shysh are former employees of Ring Ukraine, the Ukrainain branch of the Ring startup, which was acquired by Amazon for $1 billion in 2018. It also uses facial recognition in its product. After he resigned from Ring, Shysh visited China and was impressed by the scope of investment into surveillance tools.
According to a report by the Markets to Markets research company, the global surveillance technology market is expected to grow to nearly $75 billion by 2025.
Upon arrival from China, where people often hide their faces behind medical masks to cope with air pollution, Shysh decided to create a surveillance technology that can identify people beyond simple facial recognition.
That idea brought Yurchuk and Shysh to Y Combinator, one of the world’s best startup accelerators based in Silicon Valley, which jump-started their startup. Among Y Combinator alumni are Airbnb, Reddit, Yahoo, Twitch, Dropbox and other so-called unicorns — startups worth over $1 billion.
According to the co-founders, they received their first investment money after five months of intensive work. The company secures a non-disclosure agreement with clients, but Shysh told the Kyiv Post that among Traces.AI’s partners are video security services and marketing companies that evaluate the response of consumers to outdoor advertisements.
The technology developed by the two Ukrainians has already been used by clients in the U.S. and now the company is targeting Japan and the United Kingdom, the western country with the highest number of surveillance cameras per citizen.
During the pandemic, Traces.AI has considered India for testing the contact-tracing technology. With 62,800 infected and 2,100 dead as of May 11, India has kept its 1.3 billion people under the strictest quarantine in the world.
But the company is still negotiating the conditions. According to the founders, Indian authorities are reluctant to upload videos from their CCTV cameras to the servers used by Traces.AI, so the company is discussing the possibility of using Indian databases.
Potential growth
As for now, all seven employees of Traces.AI work from Silicon Valley, while mulling over rollouts in Ukraine and the European Union.
The co-founders say that now many cities in Europe are turning to cameras as a source of valuable data to help create a smart schedule for public transport, eliminate traffic jams and resolve other local problems.
Although private businesses respond faster to surveillance solutions than governments, the founders expect that the pandemic will change that.
As for Ukraine, when the prices for surveillance cameras decrease, Shysh predicts there will be substantially more of them in Ukraine, too.
“The work of our system in airports, public transport, shopping malls and on the street is crucial to fast response, as it will curb the spread of similar epidemics in the future,” he said.
