Video is the world’s largest generator of data, created every day by over 500 million cameras worldwide. That number is slated to double by 2020. The potential there, if we could actually analyze the data, is off the charts. It’s data from government property and public transit, commercial buildings, roadways, traffic stops, retail locations, and more.
The result would be what NVIDIA calls AI Cities, a thinking robot, with a billion eyes watching our infrastructure to help keep people safe.
“Historically, video data has always been used in a forensic, after-the-fact kind of use case,” says Milind Naphade, CTO of AI City at NVIDIA and one of the speakers at VB Summit: Riding the AI Wave on October 23 & 24 in Berkeley.” But these omnipresent sensors can impact everything from public safety, traffic, and parking management to law enforcement and city services.”
The challenge up to now is not just that it’s difficult to move this data, store it, and analyze with any kind of timeliness. Video is also its own special kind of creature, in the world of sensors, not like a temperature sensor or a pressure sensor that gives you just one particular indicator. Video requires interpretation via powerful deep-learned algorithms, and the kind of computational power that would allow this algorithm to operate in the kind of time that it needs for that insight to matter is massive.
“The quintessential breakthrough is that we finally have access from the edge to the cloud,” Naphade says.
Unveiled in May, Metropolis is an edge-to-cloud video platform that includes tools, technologies, and support to build smarter, faster AI-powered applications. It’s designed to put AI behind every camera, on-premises video recorder and server, and in the cloud. As neural networks are trained on increasingly complex recognition tasks, their accuracy and scalability grow to tremendous heights — and then they’re set loose to save both lives and dollars.
On a large transportation authority network, sparsely populated subway or train stations can be monitored 24/7 to summon aid for riders who encounter trouble or danger at a station — the commuter who trips at the top of the escalator, the kid who gets too close to the edge of the platform. Train tracks are subject to wear and tear over millions of miles of back and forth travel; there are more than 600,000 bridges in the United States alone, and every damage inspection causes traffic to back up for miles. Inspections by video-enabled drones would eliminate the kind of disruption that closes down the Golden Gate Bridge.
More than 50 NVIDIA AI city partner companies are already providing products and applications that use deep learning on GPUs, among them industry leaders like Avigilon, Dahua, Hanwha Techwin, Hikvision, Alibaba, Huawei, and Milestone.
With Metropolis, Hikvision has achieved recall rates of more than 90 percent for its identification and matching technology, which makes it easier to find lost people in crowded places. It works with a camera and network video recorder, plus compute-intensive system at the edge, cloud servers, and an AI supercomputer for training.
Alibaba Cloud’s City Brain offers real-time traffic management and prediction, city services and smarter drainage systems. In Hangzhou’s pilot district City Brain helped to ease traffic congestion by 11 percent.
Huawei is combating traffic congestion using intelligent video analytics, combined all the data necessary, including vehicle information, speed, direction, and more, to provide real-time traffic analysis and improve traffic flow. They have seen speed congestion rates drop by 15 percent.
And development is speeding up with their partner program, which gathers together a dozen software partners to offer a curated list of applications that make it easy for systems integrators and hardware vendors to build new products.
Among them is a facial recognition solution from SenseTime, designed for public safety, retail, and access control. The company is already working with Chinese industry leaders, including China Mobile Communications Corp, China UnionPay and Sina Weibo Corp, to leverage its technology for security and surveillance, finance, education and robotics.
Pilot projects are running in the Bay Area, where the NVIDIA Metropolis platform is being leveraged to make parking an easier, frictionless experience.
“The proof is in the pudding,” Naphade says. “We’re seeing a proliferation of the cities that are now interested in having this technology at their fingertips.”
Years ago you couldn’t live without electricity. Today, you can’t live without internet, Naphade says. And in the next few years, we’ll see AI becoming that pervasive in our lives,
“AI is going to be permanently, irreversibly changing the paradigm in transportation and cities,” he adds. “I can tell you that every city will be leveraging AI, not just for video sensing and intelligence, from edge to cloud, but you will have AI in sidewalks, AI in self-driving cars, bridges, buildings, bikes, traffic signals and more. Pervasive, right? Because this will deliver value to citizens. They’ll come to expect it. It will not be the exception. It will be the norm.”
VentureBeatVentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative technology and transact. Our site delivers essential information on data technologies and strategies to guide you as you lead your organizations. We invite you to become a member of our community, to access:
- up-to-date information on the subjects of interest to you
- our newsletters
- gated thought-leader content and discounted access to our prized events, such as Transform
- networking features, and more