Bellevue has a new ally in the battle to reduce pedestrian fatalities to zero: big data.  The city, along with Seattle and several others, are piloting new a program that uses machine learning to proactively improve bike and pedestrian safety.  The hope is that the machines, trained by a crowdsourced group of humans to recognize bikes and people, can figure out where crashes are likely to occur.

Today, planners often wait for a crash to occur before attempting to make it safer.  For example, the Rainier Ave road safety project started after it became clear that the road had an above-normal accident rate.  To achieve the Vision Zero goal will require looking not just at intersections with high collision rates, but also places where near-collisions occur frequently.

“Traffic engineers tend to be reactive,” Franz Loewenherz, Principal Transportation Planner for the City of Bellevue told me recently.  “But we don’t need to wait for crashes to occur.”  Loewenherz and other engineers are applying the idea of the “near miss”–  which is used in other industries as a safety measurement — to the urban environment.  The goal, described in this ITE Journal Article, is to shift from a reactive posture to a proactive one.

So how do you catalog a near miss?  Here’s how it works: video data from traffic cameras is fed into a cloud-based software service, in this case powered by Microsoft, a partner in the pilot project, where it’s analyzed by computers and a model is produced of how many near misses and close calls happen between bikes, pedestrians, and cars.  The software can then identify problematic intersections for the traffic engineers to fix.

Once the fix is in place, engineers can study its effectiveness.  “Was the PBL effective? Did near misses drop and bike volumes increase?  We can then share that with other institutions. ” Loewenherz told me.

The catch?  They need to train the machines to identify pedestrians and cyclists.  Cars are regular in shape and easy to identify.  But the computer doesn’t always know what to do with, say, a recumbent bicycle.  To fix this, they are opening up the program for volunteers to come in and help train the network.

This work is part of a multi-city, multi-organizational partnership called Video Analytics Towards Vision Zero. The partnership invites public participation in the next project milestone – using crowdsourcing to analyze video and teach computers to identify a person in a wheelchair, on a bike or in a car, as well as patterns of movement in intersections. The more volunteers who take part, the better computers will learn to recognize near-miss collisions.

If you want learn more, or to help teach the system how to identify bikes and peds, visit the program website.

42 Replies to “Bellevue Leverages Big Data to Get to Vision Zero”

  1. One of the biggest improvements to pedestrian and bike safety needs to be in the form of education for pedestrians and bicyclists. Three times this week alone I’ve had pedestrians start crossing the street (one not in a crosswalk) without even looking either direction for cars, with headphones on. Each week bikes also run red lights passed me, which is both illegal and dumb. It’s no wonder there are accidents involving non-auto users, just out of mere stupidity.

    Sadly, big data won’t help stupid, I’m sorry.

      1. Brings back memories, Steve, and also a cautionary lesson. About the time DSTT opened, Metro management suddenly fashioned their consciousness into walking the streets, hallways, and Tunnel platforms wearing black raincoats. Starting at sunset on a dark rainy Halloween.

        It’s true that graphic comics reveal that some part-time bats and wolves with Austrian titles of nobility are actually forces for good who fight zombies.

        But who except everybody in ATU Local 587 knows what Evil Lurked in the Hearts of innocent transit operators crossing a rain-soaked November intersection at pm rush:

        “I really, really, really DIDN’T SEE THEM!” Budget doesn’t give base chiefs lie detectors, but the needle wouldn’t quiver. We always suspected that those ridiculous personality tests we all had to take to go full-time were really to detect covert vampire killers among us.

        But cutting to the chase, torches, peasants and all: Color black on anybody official shows intent to intimidate. And wearer knows it. Along with everybody over whom they have authority.

        For their own professional honor, and survival, no police officer in the United States of America should ever face our public in a black uniform. Because whether you remember World War II German law enforcement or not, the brass skull in the hat brim is always understood.

        Wherever in Hell this fashion ever got started, back to Perdition it needs to go instead of the precinct laundry. Precisely because the lives of decent police officers matter so much, a sober shade of blue excellently represents the authority their profession deserves to carry.

        I think Metro management changed their ways when the large black hat of one of our supervisors sent children screaming down the platform for fear of being turned into gingerbread.

        Mark Dublin

    1. We need education on both sides. I see drivers sitting in bike lanes all the time. I see drivers not stopping to allow pedestrians to pass at unmarked crosswalks. I see cars crossing the bike lane without turning on their blinkers. We all share the road and we all share the responsibility of keeping one another safe. The status quo, which is an us vs them mentality from both drivers AND pedestrians, will get us nowhere.

      1. unmarked crosswalk strikes me as an oxymoron.

        Bottom-line, though, is that if a pedestrian has not established eye contact with a driver and elicited a response (i.e. – braking), he/she needs to be very vigilant in asserting right-of-way. It is the pedestrian’s life at stake, so that is where the ultimate responsibility must lie.

      2. In that case you should re-take the drivers license test, because unmarked crosswalks are a feature of Washington State law that you’re required to be familiar with.

      3. @Kevin22

        Under Washington state law, all intersections are crosswalks unless otherwise signed, and all cars within or turning onto one lane of the half of the roadway a pedestrian is crossing must stop and allow the pedestrian to cross.

        Secondly, walking is one of the things that makes cities great and it should be easy and pleasant. Walking is far more vital to a city’s health than drivers’ desire to barrel through streets with maximal speed and minimal care. You’re turning things around and putting all the cognitive and procedural burden on pedestrians- even when they have the right of way- rather than demanding drivers actively avoid injuring and killing pedestrians. The endpoint of your line of thinking is that walking anywhere becomes so laborious and stress-laden that it ceases to be a pleasant or practical way to get around.

      4. I have adopted the personal policy of never crossing an arterial, no matter how “minor”, anywhere other than a corner. My closest minor arterial is 9th NW, which I cross at one of three points depending on my destination and whether I take a detour through the nearby park/school grounds (two, one elementary the other “middle”) on the way or not.

        It’s interesting to observe drivers’ responses. Nearly all stop reliably at the two crossings by the schools. But at the one I use most, in between the other two, it’s very clear that there are two groups of drivers: those who have read the Washington Drivers’ Manual and those who have not.

        The former see me waiting and stick, the others barrel along in blissful, albeit deadly, ignorance. Unfortunately, there are no markings on the cars either group drives, so there is always an awkwardness when a car slows. Is it going to stop? Or is it going to turn through me if I start to cross?

        Hence, I try to stand far back from the intersection so that the “good” drivers don’t see me, and wait for a sufficiently large space between cars that they don’t have to do anything except not speed up egregiously.

        I never cross a major arterial anywhere but at a light or at 9AM on a Sunday morning.

        These are rules I’d urge all walkers to adopt. If you’re dead, even if it’s the driver’s fault, you won’t be able to enjoy the insurance proceeds.

    2. How do you establish eye contact with a driver when they have illegal window tint, or are driving a lifted pickup, or are staring at a cell phone? The statistics are quite clear that drivers are at fault in the majority of pedestrian fatality cases.

      1. “Establish eye contact and wait for them to stop” and you simply wouldn’t be able to get anywhere on foot. Ever actually tried this? How often do people make eye contact with strangers from 2-3 feet away while walking? Can I even see your eyes from 50 feet away, which a car would traverse in seconds, and somehow be able to see your brake lights from the FRONT of your car? Are we all Marvel superheros or something? Clearly that individual doesn’t do much walking in the city. Nor unfortunately do decision makers and even law enforcement officers.

        Driver education really should have a mandatory field walking and biking component so drivers appreciate from a young age are educated on how things look and feel from the other side of the windshield. Driving is not just operating a machine it is interacting with and sharing the road with other HUMANS. It is constantly monitoring your surroundings and making adjustments accordingly–including the 7th Commandment of not killing people. But driver licensing standards are mainly based on machine operations and distracted driving is still a secondary offense.

      2. Or they’re turning left onto the road you’re trying to cross, and are too far away for you to see their faces.

    3. This week, I’ve had a half-dozen near misses by cars (both as bike and pedestrian), where I had complete and uncontested right-of-way and about half of those were turning with no signal. I only count a near miss where I had to make a split decision to AVOID getting run over. That doesn’t include the countless times where someone drives aggressively around me, but I don’t need to avoid getting run over.

      Is it obliviousness on the drivers part? Stupidity? Privilege? I’ll probably never know until I get run over and sue the ever-living daylight out of them, but it’s clear that drivers need just as much education as anyone else. I do know that the “war on cars” isn’t going very well for the non-cars. I haven’t even come close to claiming my first car kill as a cyclist or pedestrian :/

      1. I spend a lot of time in San Francisco. There, I do A LOT of walking and bicycling. Here I bicycle A LOT, but don’t walk much. But, yes, while I may enter into an intersection when a car is approaching, I stare it down, and if I can’t establish eye contact then I don’t proceed. I succeed almost all of the time. To say it’s not possible is to say you are blind or that you are trying to cross streets like Rainier at “unmarked crosswalks”, rather than at the corners where the signal lights are. It’s just common sense. You can prattle on all you want about “burden”, be bitter, etc. but it’s your life at stake. And if you insist on being bitter, then why not agitate for this city to get off its cheap, lazy ass and paint more of the crosswalks and/or build pedestrian bridges. Would be a lot cheaper than “road dieting” all the streets.

      2. Building pedestrian overpasses all over the city would be significantly more expensive than road diets, which are mostly just paint.

      3. @Kevin22: I also do not move in front of a car’s path unless I can see them looking at me. That doesn’t stop someone from flooring it through a stop sign or flying around the corner with no warning.

        Self preservation is my goal, I have no intent on getting hit to make a statement; your post almost has some tones of victim blaming in it. But act as smug as you want and try to paint me as an oblivious pedestrian, but you couldn’t be farther from the truth.

        That still doesn’t excuse aggressive drivers ignoring the law and putting our lives at stake so they can get to work 5 seconds faster.

    4. Sorry, but while your observations are correct and you have a point, the data shows that that is NOT the reason behind cyclist and pedestrian fatalities and near misses. Fatalities are mostly due to right and left turning cars failing to yield at intersections and for bicyclists additionally the “door zone.” And these crashes happen regardless of cell phone or headphone use on the part of the victims. Banning right turns on red eliminates many of these conflicts as well as T-bone vehicle crashes.

  2. I understand the jump to “whiz bang” gets articles published by ITE, but I doubt any of this is better than a simple ban on right turn on red in downtown environments.

    1. Exactly.

      We already know what and where the problems are—the issue is that they’re *politically* difficult to solve. Using “Big Data” is just an buzz-worthy excuse to keep studying the problem and avoid the political difficulties of actually implementing the solutions.

      You hear this kind of rhetoric with global warming as well: “the science is unclear, we need to keep studying the problem”.

    2. YES.

      This article breathlessly talks about being less “reactive” so the answer is… machine-learning that needs more data-collection to recognize people walking and biking?

      The simple fact is that the safety and convenience of pedestrians (which includes essentially all transit users on each end of the trip) and cyclists will often come into conflict with motor-vehicle capacity in well-known ways. If we don’t stop prioritizing motor-vehicle capacity every time we won’t achieve our pedestrian, bike, and transit goals. This doesn’t call for more magic machine-learning, it calls for old-fashioned political will.

      1. You interpreted his quote differently than I did. “Traffic engineers tend to be reactive,” To me that is just a statement of fact. There are other things they do from a safety perspective (study possible changes, adhere to modern codes when making changes, etc) but the most common way they improve safety is to react to past failure. That is really not that different than a lot of safety science. With jet aircraft they spend a lot of time imagining what could happen, but they also spend a lot of time looking at what happened. But unlike traffic safety, they spend a lot of time looking at close calls. Other than anecdotally replying to a complaint (“Hey, fix that intersection, I almost got killed”) traffic engineers don’t do that. Or at least, they haven’t in the past. Because unlike air travel, the number of close calls is huge and hard to track. Until now. That is the point of the article.

        Of course we need more political will when it comes to prioritizing safety, but we are certainly moving in the right direction. Speed limits have been lowered all across this city. There are particular intersections that need safety improvements, but it isn’t always cheap and easy to figure out how to do it. We need to prioritize and fix the most dangerous ones first, don’t you think? This simply allows the engineers to do a better job of prioritizing — of fixing dangerous areas before the number of accidents becomes so high it is obvious they need to do something.

      2. Exactly.
        Bellevue has data on collisions.
        Bellevue is aware that the majority of the city is unsafe for walkers and bikers.

        Bellevue is ignoring all that data, and embarking on a re-visioning effort that’ll delay any safety fixes for a few more years.

        Northrup way is still broken, at least six months late, delayed building inadequate and unsafe bike infrastructure, and there are still no alternate paths.

      3. And now they’re gonna go install a bunch of cameras at intersections instead of missing crosswalks and bike lanes. SMDH. If they have to walk just a couple blocks to do their jobs they could collect all the data they need right there in person. If they manage to accomplish their jobs without walking, they could just ask people that have ever walked in Bellevue.

        It doesn’t take that much data. You can literally ask anyone. Installing cameras and training an f’ing machine learning model is a total waste of time — it’s a lot of work for humans to determine the underlying truths uncovered by machine-learning models, and in this case that’s necessary because the ultimate result has to be human-created road designs. What it does take is a lot of is will. The problems, experienced by people that do walk in Bellevue (and Seattle), are known; the solutions, which have been implemented elsewhere in the world, are known; the obstacles to their implementation in Bellevue (and Seattle) are known. Leaders would rather hide behind “smart” technology — “Sorry about those long waits to cross the street, the computer made me do it!”, says SDOT regarding Mercer — than address the underlying challenge, which is a political challenge caused by our choices, head-on.

  3. Bellevue is a terribly pedestrian unfriendly downtown. I jaywalk all the time, simply because I don’t want to put up with waiting 2 mins at every light…

      1. It it really were two minutes at every block, I suspect that’s exactly what would happen. Additionally consider that unlike with cars, waiting the 2 minutes does NOT generally reward you with cross traffic conflicts being eliminated–drivers are literally DIRECTED to turn in to pedestrians while they’re crossing!.

      2. B, Are you saying that Bellevue has (an) intersection(s) at which a Walk light is on for a particular direction and that there is at the same time an illuminated right turn arrow directing traffic to expect an un impeded right turn? That is so stupid that I frankly cannot believe that it is true.

        Otherwise, what do you mean by “drivers are literally DIRECTED to turn into pedestrians while they’re crossing”?

      3. I think it means cases where pedestrian crossings and car turns are the same cycle. For instance, westbound Pine Street at Boren with cars turning right, or leaving Costco in SODO with cars turning left.

      4. Closer to 3 minutes in Factoria, and yes, the cars share the cycle because there aren’t directional arrows – they have right or left turn privileges (green ball no arrow) at the same time the walk signal is showing (which is almost long enough to step off the sidewalk before it starts flashing and the cars start believing they have the right of way again). This would be specifically Factoria Boulevard at 36th, 38th, and 40th. If you have to cross both ways – to go from one corner to the opposite corner – plan on five minutes unless you time it just right.

  4. New data can help frame a problem but a human has to identify the good solutions.

    Rather than spend money on interpreting data from machines, an independent team of trained intersection designers that can go from city to city and suggest solutions could be more effective. Lots of cities have a very small staff or a culture of “do want my boss or manager or Council member wants”. It usually takes views from several people to solve a human design problem.

    Consider that I n Downtown Bellevue’s case, there is a historic design culture of narrow crosswalks at intersections. It’s hard for a driver to see a pedestrian in a narrow crosswalk that spans 5 or 7 lanes! All the studies and data in the world can’t solve this design flaw. It requires moving the lines — and the sensors in the pavement (which staff usually do not want to do).

  5. While tools like this would certainly help get us to Vision Zero, the reality is that we already have “Big Data” that show us where collisions are already happening. What we lack is the corresponding “Big Budget” to fix them all.

    For example, this cool visualization tool developed by Tim Ganter in collaboration with Seattle Neighborhood Greenways shows that since January of this year there have been 3,654 reported collisions in Seattle, involving 199 people walking or biking. The estimated economic costs are $57,025,800.

    As another example, the City’s recently published 2017 Vision Zero Progress Report [PDF] has a map (page 18) of High Crash Corridors where collisions occur “too frequently.” It’s eerily similar to an arterial map of the city.

    Before we start using shiny new tools like this one to find places where people may get hurt, I’d much rather see our scarce transportation budget focused on fixing the places where people are already being injured and killed.

    If you’re interested in making Seattle’s streets safe and comfortable for people of all ages and abilities to use, consider connecting with the Seattle Neighborhood Greenways group in your neighborhood. You’ll likely find an engaged group of like-minded neighbors to work with on projects like safer crosswalks, protected bike lanes, lower speed limits, and, of course, greenways.

  6. The entire time I knew my grandfather, a large part of his job was to drive to an intersection and count cars. His lower left arm was always tanned to brown leather from resting on the car window while the rest of him was pale as a ghost. Today that job is usually done by a metal box attached to two sensor lines.

    This new technology is currently pretty good at counting cars, and engineers tend to like to improve things that are measurable. In an imperfect world, the right thing to measure isn’t always the easy thing to measure. We measure cars, maybe we can make those numbers better. Who knows what happened to the experience of people walking, biking or riding the bus?

    This specific tool is almost certainly imperfect, but computer vision tools will likely take the place of those metal boxes that now do the job my grandfather once did. If those tools for lack of quality training don’t reliably recognize people on bikes, or in wheelchairs, or on scooters, or kids or someone with a walker, or someone dressed in black at night, or if they don’t recognize people at all, but simply cars, then we will likely continue to have engineers focused on improving the numbers for cars.

  7. On this one, The Blame Game is Getting Lame. Of course skillful driving, good manners, and common sense keep everything safer than without them.

    But The basic mechanics of driving an automobile make these machines incompatible with the streets of a city the size of either Seattle or Bellevue.

    Shouldn’t be necessary to forbid them. Just see to it they keep the posted speed limit of ten miles an hour cross streets, and fifteen on arterials. And stay out of transit and bicycle lanes.

    And limit parking to structures at transit terminals. And provide bus, car-share, and taxi service a five minute walk from average regional door.

    Nobody puts Lyfft on grooved pavement in front of LINK. Or pedestrians on a plaza that’s also a freeway. With present numbers of cars, pedestrians, and bicycles, mixed traffic will inevitably kill people traveling last two ways.


  8. What Rainier Ave safety project? If actual injuries and deaths don’t make it politically feasible to make a street safer why would ai data about near misses be taken seriously.

    1. It has been in discussion for a few years after a car crashed into a retain shop in Columbia city and later another car did the same thing. City stats show that mid Rainier has has among the largest number of collisions and fatalities in the city, and that a reconfiguration of Rainier Ave and MLK Way could bring that significantly down. They pointed to the high speed limits and expressway-like nature of Rainier and the complicated intersection at Rainier, MLK, and McClellan. They proposed various alternatives and I don’t know what won out, but one of them was a bow tie that would eliminate half the turms. The Rainier problem was one of the buggest impetuses behind Vision Zero, which seeks to reduce collision fatalities to zero through lower speed limits, road diets, and ped/bike/bus access improvements. That was partly in the Move Seattle levy with its “Safe Routes to School” program.

      It also intersects with SDOT’s and Metro’s plans for the 7. They plan to split it at Mt Baker, attach the south part to the 48, and make both routes RapidRide.

      1. The “split” of the 7 onto Route 48 is at 23rd, not directly at Mt Baker. That means that the buses will have to jog twice (left and right turns 100 feet from each other) at Mt Baker Ststion if the City’s last MLK/Rainier proposed reconfiguration is executed.

        That’s not an improved “intersection” for bus routes.

  9. So maybe this program is a bit ahead of where it needs to be, but Bellevue is trying to achieve Vision Zero ( And if the data on the Crash Map linked from that page can be believed, Bellevue is nowhere near as dangerous as its reputation.

    I’m a bit disappointed at the hostility in the comments here. Yes, Bellevue has a bunch of obvious problems when it comes to walkability. The city is working on them, but it takes time to change from an automobile-only city (as it was on a couple decades back) to a pedestrian-friendly urban core. Should they not invest in any long-term research projects until every person on STB believes that they’ve fixed all the other problems in the city?

  10. Big data can’t help Bellevue.

    The wide streets and superblocks are the problem. (Note that Main Street in Bellevue is an exception — with a narrower street and smaller blocks.) Does any pedestrian feel safe crossing 6+ lanes?

    1. Yes, I cross one in Bellevue every day as I walk from my office to lunch, along with about 20 other people during the lunch rush. Never felt unsafe.

Comments are closed.