If you ride on NE 26th Avenue and Wasco, you might have noticed a small square box with a camera lens peering out onto the street. The box is attached to a garage and accompanied by a small LED display that shows the number of bikes, pedestrians, and cars that have gone by. This isn’t just a trivial weekend project by a garage tinkerer, it’s the result of years of development by a data expert who thinks his creation can revolutionize urban planning and kickstart a new era of people-powered advocacy for healthier cities.
The device is the work of Matt Zajack, a Sullivan’s Gulch resident and data analytics consultant who’s about to launch his new product into the world. Zajack’s Traffic Monitor is an, “open source roadway object detection and radar speed monitoring,” device made from simple and widely available components. It uses a wide-angle camera, a very small (Raspberry Pi) computer, and machine learning and artificial intelligence (AI) code to watch for objects in the street and then tallies what it sees in a database and presents the information in an easy-to-read display. Add a few sensors — for things like air quality, temperature, and doppler radar to measure speed — and you’ve got a trove of useful data in a small package that can be deployed almost anywhere.
“I am talking about capturing the picture of the roadway biome. Trying to get people to think about: How healthy is your street environment? Does it encourage, promote, support human movement? How can it adapt to changing conditions and environments?”
– Matt Zajack, Traffic Monitor
I first met Zajack and learned about this project in January 2024 when he signed up for a 15-minute slot in my weekly office hours. I was so intrigued, I encouraged him to bring the device to Bike Happy Hour. The night he showed up, I recall many people being amazed by it and a curious crowd that had gathered around him. Since then, Zajack has continued to develop the product and now that it’s nearly ready for prime-time, I rolled over to his garage on Monday for a closer look.
As a 3-D printer whirred in the background, I learned more about Zajack and his vision for the Traffic Monitor.
“A couple of years ago, I realized there are a lot of organizations, like Bike Loud PDX and others, who were looking to measure the impact of the programs they were doing, and also argue against a lot of the the sparse data that was out there that was only really provided by municipalities,” Zajack shared. “A lot of the only solutions for traffic monitoring cameras are from private companies doing implementations that cost tens and hundreds of thousands of dollars that can really only be used by municipalities and very large companies.” That’s when Zajack, who wants to use “AI for good” (according to his consulting website), realized, “We can do these things for a lot cheaper.”
These “things” include a number of valuable use cases. Consider a neighborhood association that’s worried about cut through traffic or speeding on a street. They see traffic calming as a remedy, but need to make their case to the Portland Bureau of Transportation and/or their city councilors. The most important part of that case is often data: Can the neighbors prove people are speeding? Can they prove there are more cars on the street then PBOT guidelines call for? If you’ve ever tried to get that data using City of Portland’s existing sources, you know how frustrating that can be. If they have traffic counts at all, they’ll only be cars and they might not be in the precise location you need or they might be many years old. Want bike counts? You’d have to convince PBOT to set out a pneumatic hose counter. Want pedestrian counts? Forget about it.
Zajack says his primary selling point is that data from a Traffic Monitor will be much more robust than anything currently available, it will provide recent data, and it you can move it to any location you want (thanks to 3-D printed cases with tripod mounts).
A Traffic Monitor that includes an air quality sensor, doppler radar (for measuring vehicle speed), and two cameras costs about $400 to build. Once installed and set up, the device can detect just about anything that happens on a street — not just the number of cars or bikes. Zajack says the benefit of his device is that it gets a “holistic picture of the roadway.” “What I have here is a wide angle lens that look sat both sides of the sidewalk and the roadway. So we’re counting pedestrians, bikes, cars, I also added in dogs just for fun, anything that goes on on a roadway.”
Zajack elaborated on this idea of getting a holistic picture of what’s happening on the road: “I am talking about capturing the picture of the roadway biome. Trying to get people to think about: How healthy is your street environment? Does it encourage, promote, support human movement? What is the mix of users? How do people get around? How can it adapt to changing conditions and environments? The roadway is used by more than just, you know, 5,000 pound steel cages moving through it.”
And if all this monitoring makes you queasy from a privacy perspective, Zajack’s unit does all the processing locally. That is, nothing is sent to a cloud or a third-party server. All images used are deleted a few seconds after being taken and owners of the Traffic Monitor will be the only ones with the information. Of course, there’s no way of enforcing this type of privacy protection, but Zajack says his device is no different in that respect to any other type of widely available security camera.
“We can crowdsource this. We don’t have to send this to a large, private corporate cloud. We can do this the processing locally. We don’t have to have all of the decisions being made from a central location,” Zajack continued. He sees his role not just as developer and consultant, but as a teacher of how technology can empower a community in an ethical and transparent way, instead of being forced onto them by entities beyond their control.
The next step would be having the City of Portland either accepting the data from individual citizens, or deploying the devices themselves. He’s also worked with nonprofit group Portlanders for Clean Air, who bought one of his first units to measure diesel engine particulate matter from large trucks.
As chair of his neighborhood association and active member of Bike Loud PDX, this product is also personal for Zajack, who purchased his home in 2020 because it was on the corner of a neighborhood greenway route. He thinks planning decisions are often unfair to people who don’t drive because walkers, bikers, scooterers, joggers, and so on are invisible when it comes to data and traffic counts.
By the time I had met Zajack around noon on wet and cold Monday, the LED display he built for the Traffic Monitor attached to his garage read: “Bike: 78; Ped: 144; Car: 242.”
“When someone from the city is like, ‘Oh, we’re not going to tackle that problem because there aren’t enough people walking. There aren’t enough people biking.’ I’m like, that can’t be true! I spend a lot of my day on a greenway, and there’s a lot of lot of human activity here that isn’t centered around cars that’s not impacting the decisions that we’re making. And it’s a real shame. So I kind of want to change that equation.”
Zajack’s ideal vision is widespread deployment of the devices and individual Portlanders working together to crowdsource data to hasten planning decisions on their streets. Using cheap technology to fight larger, systemic power structures, is something that animates Zajack. “There’s a reason I have it open source, low cost, and low barrier to entry — so anyone can build them,” he says. “I want everyone to see the power of what we can do with a very low price point, with the tech that’s available today to anyone.”
“I want to democratize the power of AI, to get it into the hands of the people.”
Meet Matt and see the Traffic Monitor in person at Bike Happy Hour this week, 3-6:00 pm at Migration Brewing on N Williams Ave.
Thanks for reading.
BikePortland has served this community with independent community journalism since 2005. We rely on subscriptions from readers like you to survive. Your financial support is vital in keeping this valuable resource alive and well.
Please subscribe today to strengthen and expand our work.
Wow! This is really great. The data that has often been pointed to, by PBOT and others, to determined street design has always seemed to be too simplistic. Particularly for greenways, there are streets that feel relaxed and safe, and others where I have routinely had to deal with reckless drivers. The car counts seem too simplistic. I don’t know if PBOT routinely measures car speeds as part of their evaluations. And I imagine that PBOT probably does not have a way of capturing cars rolling stop signs and having conflicts with bikers on greenways. Ankeny and Tilamook come to mind.
Going St. is worse than either of those.
I could see that. I don’t ride it often, but when I have, I’ve noticed some problematic areas. NE 7th, which seems to be a half-hearted greenway used as a car sewer overflow from MLK, is one of the worst in the area for me.
7th was supposed to be a greenway, but PBOT backed off on putting in diverters and traffic calming after some folks were upset about losing car access.
Do you need to train it for awhile after install?
By the way, we all appreciate the misting tunnel you setup in front of your garage on the hot days! And sorry about your heritage tree.
No, it needs no training
If the city doesn’t implement these I hope some local groups get together to do it. I would donate money to BikeLoud, my neighborhood association, Street Trust or whoever else wants to implement these around town.
I’d also be interested in having one on my corner more because I’m a data nerd and like this sort of stuff but also because maybe I could capture enough speeders to get the city to install a stop sign at my uncontrolled intersection. There’s a ton of people walking through that intersection and no sidewalks so speeders are a big safety issue. The landscaper in the backhoe is of particular concern.
The data is aggregated locally, but does that mean it also needs to be downloaded manually? It’d be cool if we could get a network of these set up that uploaded to a common website, sort of like how Weather Underground relies on a lot of people’s backyard weather monitors.
Good Q’s, Will. The data is event-level (no aggregation) with ability to 1) connect and upload (de-identified data) to an open source IoT hub, ThingsBoard.io, of your choosing or self-hosted 2) download in SQLite format for local analysis, 3) subscribe to your device MQTT feed for live data to another system you set up. All that in addition to the basic on-device dashboard (for end-users).
And a network of them is in my mind. That is separate from the open source traffic monitor software project. I do believe we can crowdsource the data for more powerful decision making, whether it’s for a specific initiative or longitudinal analysis. I am getting close to hosting a platform with number 1 and ask folks to contribute, just like many do to PurpleAir, Wunderground, FlightAware, etc. as you mentioned.
It will be interesting to see if PBOT will accept the data produced at face value, and whether it makes it easier to get them to take your requests/suggestions/complaints seriously.
it will also be interesting to see if City Council will accept the data and provide new policies and guidance to the city administrator and deputy city administrators and direct/compel PBOT to act
This is cool! I’ve been wondering how hard it would be to make something like this for a while, but I didn’t feel like I had the right skill set.
SUPERB !! the revolution begins today
The road to hell is paved with good inventions.
Recent experience has shown that no one needs to feign good intentions (?? although I like the play on words with inventions) to bring you down the road to hell. That said, I do (honestly) appreciate your healthy skepticism!
In fact, to your point: I hope this, as a glass box system, educates everyone on the inner workings of computer vision, AI, and surveillance systems. That way we can better monitor our institutions, demand AI transparency (responsible AI), and create regulation concerning its use, much like the EU AI Act.
To give you a local, PDX example that I am making up: When PBOT installs speed safety cameras (license plate cams), we should demand that whatever vendor they choose is able to fill out a “scorecard” or “AI Transparency Report” created by, say, a Transportation Community Advisory Committee, that includes things like: how it works, what models it uses, how those models were trained, where the data goes, what the data includes, how the systems can be audited, what the data retention definitions are along the way, and more!
what an accomplishment! bravo!!