Alibaba linked up a city’s worth of security cameras with AI that sees problems before they happen
What do security cameras and “a group of blind people trying to describe an elephant” have in common? According to Alibaba’s Distinguished Engineer Dr. Hua Xiansheng, this ancient Indian parable in which people grope for a sense of the whole animal is an apt metaphor for what you get when smart machines can only process what is right in front of them.
Far from a lament of technology’s shortcomings, this “forest for the trees” problem is how Dr. Hua recently framed a new Alibaba AI venture, known as “City Brain”, for audiences at The Computing Conference in Nanjing. Built to make smarter use of aging camera infrastructures, City Brain coordinates data across every camera in an urban security system to enhance the smart functionality of each node and deliver actionable insights, from traffic forecasting to accident response. And with one of City Brain’s three component products now deployed throughout Hangzhou city, the proverbial blind elephant-survey is already giving way to vision-enhanced digital traffic administration for a population of almost 10 million.
Crisis Response and Prevention
City Brain’s first component, TianYao, couples 24-hour pedestrian and vehicle traffic monitoring with algorithms that recognize anomalies and notify authorities of incidents like traffic congestion and vehicle collisions. It also processes image feeds over time to identify sites with high incident frequency, helping human administrators locate problems and focus solutions for maximum impact. Adding to its sophistication, TianYao then offers immediate feedback on the effectiveness of those solutions. In practice, Tianyao has already proven an effective replacement for traffic patrol officers in Hangzhou even under low-light conditions, freeing up human labor to address more urgent tasks and emergencies.
With City Brain’s second TianYing component, Alibaba introduces a technology more effective at pedestrian recognition than human agents provided with the same visual feed. TianYing draws on Alibaba’s two decades of image recognition research, specifically focusing on the concept of an index object (such as a person of interest) and visual cues that can describe and confirm that object in a camera scan. With 96% recognition success over the human average of 94%, it offers a powerful tool for quickly and accurately identifying missing persons or tracking hit-and-run fugitives, for example, freeing human agents to focus on pursuit and apprehension.
Paving the Way for Smarter Traffic Management
City Brain’s last component TianJi, still in development, is a predictive analysis tool for vehicle and pedestrian traffic flows. With early testing in Suzhou city confirming it can accurately predict traffic patterns in time spans ranging from minutes to an hour, TianJi offers the hope of better traffic management and could soon enable early warnings to help individual commuters plan their travels.
As one of the first Four Open Innovation Platforms for AI chosen by China’s Ministry of Science and Technology, the majority of City Brain is currently open for public contributions to development. With early implementation standing up to the demands of real urban settings, Alibaba anticipates City Brain and other AI systems will prove an indispensable urban resource comparable to utilities and physical infrastructure in the near future.
Know more about Alibaba’s CIty Brain: