DeepGlint wants to make every CCTV camera searchable

If you buy something using links in our stories, we may earn a commission. Learn more.

This article was first published in the April 2016 issue of WIRED magazine. Be the first to read WIRED's articles in print before they're posted online, and get your hands on loads of additional content by subscribing online. For more stories from WIRED's China issue, click here.

A poster of London buses hangs on the wall at DeepGlint's offices, in north-west Beijing. "London has the most video surveillance cameras for its population - 2.2 million, according to my data - but who's watching these cameras?" asks Bofei He, the company's 35-year-old CEO. "Nobody. Human beings can't watch 50 screens at a time. Or if you watch, you don't see."

DeepGlint wants to solve that problem by building a visual sensor network that understands what's happening around it. "What if we had sensors to read what's happening everywhere simultaneously, and digitising that world?" He asks. "We want to enable computers to understand the world as well as humans can."

Its solution: a depth-sensing camera system that makes the physical world searchable. Some of its first customers are banks: "In the past, you needed two people present when an ATM was being loaded in case counterfeit currency was being put in. We can ensure they're following procedure, and can keep customers safe - our system notices if someone withdrawing cash is being approached."

The 80-person company is also working on vision systems for autonomous cars that He says offers a $500 alternative to Lidar; it's spun off a separate company, UISEE and its goal is "to work more closely with Chinese municipalities". "Half of all hard drives are sold to the surveillance industry to store data," He says. "We're talking with the Tiananmen Square municipality which has cameras everywhere, generating 1,800TB of video data every day. If you burn that 1,800TB on to DVDs, every 24 hours that pile is taller than the Eiffel Tower. Good luck searching for certain scenes in that movie." At airports, he says, the system can be tweaked to identify unattended luggage, or passengers moving wrongly through security checks. "And we have engineers working on facial recognition."

Surveillance and crowd-tracking: this does sound like a particularly Chinese set of priorities. But He is sensitive to how his company may appear to a westerner. "We are not a surveillance company," he says, sitting by a lake in what used to be a country club for high-level government officials, right beside the Summer Palace (an anti-corruption drive forced the country club to offer workspace to startups). "We're a computer vision and AI company. And an independent company, with only independent investors." These include Sequoia Capital, ZhenFund and Ceyuan Ventures.

Besides, DeepGlint has Silicon Valley origins: He's co-founder, Yong Zhao, helped start the Google Glass project when he was at Google, after spending six years at Brown University; and He studied at Stanford. "We understand that if ever we want to go global or go public in the US, we'll need to answer questions as to our relationship with government here," He says. "We want to be clear from day one that we're independent - we treat them as a customer."

This article was originally published by WIRED UK