What if computers could help prevent traffic deaths before they happen?
A video analytics project called Vision Zero aims to do just that — but first it needs the help of some humans.
Volunteers are needed to help train computers for the task by watching short video clips and marking certain objects — such as pedestrians, wheelchairs and strollers as well as cars, motorcycles and buses — and flagging “near misses” at intersections.
By compiling a bunch of similar crowd-sourced data, computer programs can recognize patterns and pinpoint areas of possible trouble.
In turn, local traffic engineers can then address problems before they happen.
It’d be a big step away from the status quo, where government agencies wait until a sufficient amount of bad stuff has already happened to justify an improvement — a method that also doesn’t account for near misses or unreported incidents.
“This technology could revolutionize traffic safety analysis,” said Ryan Peterson, a transportation engineer with Snohomish County.
The tool could help better time traffic signals, manage traffic flow or target spots to place new crosswalks or other safety improvements.
“Snohomish County Public Works does not have the resources to place staff at each intersection along a corridor for hours or days to observe pedestrians and traffic to identify specific problems,” Peterson said. “The Vision Zero project could make these same observations and potentially collect this data in a much more efficient manner.”
Snohomish County is among the partners in the project. Public Works staff are donating time and traffic camera footage.
Microsoft, Bellevue and the University of Washington have led the charge on the project.
The project reaches beyond Washington state. But local governments and groups dominate the list of partners. Cascade Bicycle Club is another supporter.
Now, the instructions for volunteers aren’t as simple as point-and-click.
Volunteers are asked to mark when a pedestrian disappears behind a garbage can, for example, or when a car leaves the given study area.
You’ll be helping a computer brain develop its “deep neural network,” what project researchers describe as a machine learning system inspired by the central nervous system of animals.
Learn more and volunteer at www.ite.org/visionzero/videoanalytics.
Have a question? Email us at streetsmarts@heraldnet.com. Please include your first and last name and city of residence. Look for updates on our Street Smarts blog at www.heraldnet.com/tag/street-smarts.
Talk to us
> Give us your news tips.
> Send us a letter to the editor.
> More Herald contact information.
