After months of collaboration with Microsoft and the University of Washington, the city of Bellevue premiered its crowd-sourced video analytics project Thursday morning in the hopes of making the streets safe for everybody.
The project is a traffic-based one intended to reach the city’s goal to have zero traffic fatalities in the city. Anyone will be able to get on the project’s website to help a machine learning experience to improve traffic conditions. By pinpointing where collisions or near-collisions have occurred, the city might be able to make changes to prevent them from happening again.
If that sounds complicated, it’s because implementing artificial intelligence is still pretty untrod upon territory. Franz Loewenherz, principal transportation planner for the city, said the project was far reaching for a good reason.
“Every year, more than 1.2 million people die on our roads around the world,” he said. “Vision Zero is an international movement. Imagine a future where you don’t have to wait for crashes to occur to predict ones in the future.”
In the United States alone, 40,000 people died in 2016 and 4.6 million people were injured. For U.S. residents under the age of 19, traffic collisions were the leading cause of death. In Bellevue, 433 traffic incidents resulted in an injury, 39 bicycle crashes and 47 pedestrian incidents. One person lost their life at a downtown crosswalk last year. Vision Zero hopes to end all traffic deaths and serious injuries in traffic by 2030.
The project’s official name is Video Analytics Towards Vision Zero, and it uses researches from the University of Washington’s PacTrans transportation consortium, machine learning experts from Microsoft and Bellevue’s infrastructure to study intersections to prevent collisions before they happen.
Bellevue Mayor John Stokes said the partnership was a natural fit.
“In this case, we have the opportunity to work with great partners to make the biggest difference of all — saving lives,” he said. :Everyone who participates in the video analytics projetc can make a small contribution that could add up to a big impact in keeping our walkers, bicyclists and drivers safe.”
With the help of the public, the system will be able to identify pedestrians, vehicles, cyclists and busses and will flag “near-misses,” or a situation where a collision was imminent before one of the parties averted. Anyone can do that by watching a 10 second clip of a busy traffic intersection from a city somewhere in North America. Volunteers will draw boxes around each moving entity and trace where each one travels during the clip. By inputting hours of raw data, the computer system will learn through what is known as a Deep Neural Network, sort of like how the human brain learns.
Additionally, the program (which is almost like a game) is highly addictive.
Dr. Victor Bahl, one of the progenitors of the project and a Bellevue resident, said the project needed people to help if it was going to be successful.
“We need people to help us crowdsourcing,” he said. “With that help we’ll be able to have real-time detecting data. Near-misses are rarely reported to the police, but with that data we can create heat maps of intersections to find where traffic issues occur.”
Nearly all the speakers had a personal anecdote to go along with the presentation.
Bahl referenced several colleagues working at Microsoft’s Cambridge, Massachusetts branch who had been struck and injured while cycling. Loewenherz said that just last weekend his 11-year-old son had been struck by a vehicle as he rode his bike. Mercifully he only needed stitches, but the point remained.
For Lorraine Stewart, an avid cyclist and ride leader with Cascade Bicycle Club, it was even closer to home.
“On April 14, 2016 I was out scouting a route. It was a sunny day and I was struck from behind,” she said of collision in Bellevue which brought her to the brink of death. “I had a feeling of being crunched and then I was looking up at the sky thinking I was going to die. If there’s anything I can do to prevent just one accident, I want to be part of that.”
Stewart is out riding again, but is also helping participate in the program.
On a tangible level, what the video analytics program can help do it create change on an intersection-by-intersection basis. If, after data is collected and the machine learning is becoming more effective, it flags a certain intersection in Bellevue as having a lot of near-misses, the city could effect change. Vision Zero, an international program, has seen just this occur in Sweden and France, Bahl said.
“The data was seeing a lot of accidents and near-misses at an intersection in France,” he said. “The engineers put a roundabout there and the rate dropped to almost nothing.”
Also, as self-driving cars become more of a reality, Bahl said there might be a significant market to make the programs at the intersections and the vehicles (and even bikes) communicate with one another to make the street safer.
The program launched in several other cities on the same day, including Calgary, Los Angeles and New York City.