Traffic experts met last week at Spaces NoMA for the fourth Playing with Traffic event of Transportation Techies.
A handful presented their latest work in a rapid-fire show-and-tell of the wide array of open-source mapping and imaging that can now inform how streets are planned for both current users and future technology.
New ways to see
Mapillary’s Janine Yoong explained how combining computer vision – using digital images to train computers to understand objects – with human collaboration can inform the development of autonomous vehicles.
Janine Yoong of Mapillary
Yoong and her team hope to use street-view images from across the internet to help driverless cars better categorize items that they “see” while also creating fresher, more accurate, and complete maps that can help computers understand their location.
With this, Mapillary pulls images of streetscapes from around the world, including remote arctic research bases, that can train AV programs by processing as many objects and situations as possible. Because of the enormity of the task, they rely on volunteers to sift through segments of each image by identifying objects in one of 97 different categories that will help the computer eyes learn, much like public agencies have begun to do with traffic cameras in their Vision Zero efforts.
Creating these image-based maps can help as planners develop a better understanding of existing facilities for bikes, pedestrians, and transit users, and where they add to them while optimizing AV use to serve these modes.
In this spirit, Mikel Maron explained how Mapbox Cities uses data tools to make resident-centered decisions in fulfilling initiatives like Vision Zero, which is an effort to eliminate traffic-related deaths. By combining geographical data from sources like OpenStreetMap – in which Mapillary’s computer vision training plays a role in adding street signs – with various data points from sensors and local demographic information, Maron and his team are able to identify key intersections to study potential interventions.
Mikel Maron
Maron pointed out that an important aspect of these partnerships is to give planners a suite of information around which they can plan concrete actions that focus on the input of residents and their wellbeing.
This can build upon the work of Techies like Xiao-Feng Xie, who pulled crash data from multiple sources to better identify safety problem spots. He emphasized that tapping into more than one database, such as federal- and state-level reports, makes for more detailed and accurate analyses. Folding work like this into something like Mapbox Cities can make sure planners push useful, digestible data out to people in a way that they can respond to in a way that helps planners make streets safer and more people-focused.
Xiao-Feng Xie
Fun with transportation
Though the people who attend Techies have really embraced the joys of playing with transportation, MAGfest’s Nick Marinelli (pictured at the top) showed the group how video games have, in sometimes roundabout ways, portrayed transportation and urban planning as fun parts of pop culture. Classic games like SimCity introduce a straightforward sense of urban planning that highlights the causes and effects of different decisions, while more unexpected examples like Grand Theft Auto or Crazy Taxi exemplify simulations of low-density traffic or the evolution of navigation and pathfinding, respectively.
Another Mapbox team member, Jacob Ellena, brought some video-game pathfinding tropes into real life for augmented reality navigation. Through Unity, a video-game tool that Ellena describes as a 3D renderer, he and his team can build 3D maps of cities that place users within them through their phones. Travelers can then navigate using GPS as the AR provides visual cues like chevron arrows and sparkling waypoints to prompt a change in direction, much like in racing games.
Jacon Ellena of Mapbox
And for those who just like to sit and watch, Maxim Leyzerovich showed how easily the public can compile images from D.C.’s traffic cameras, having built a “god-mode” webpage that pulls the screens from across the city’s network. Though the project stemmed first from a desire to check traffic conditions outside his office, and then to capture glitchy photos with an artistic aesthetic, it shows promise as a true analytical tool. This project both provides insight into collecting data on roads, and from there analyzing their impact on society. This is a page that could serve as the basis for plenty of future Techies projects.
Maxim Leyzerovich
Finally, Antonio Zugaldia, also of Mapbox, showed the crowd how anybody with Android programming skills can build their own robocar, passing around his own as an example. It turns out they’re pretty easy to build with the right programming, which can train the robocars to filter objects by color as they do with guiding lines, or in object recognition and avoidance. Zugaldia’s fun-sized example carries fewer mobility implications than a full-sized passenger vehicle, but if the programming is so accessible now, perhaps it does suggest widespread AV adoption may not be far off.
Photos by M.V. Jantzen.
The post Traffic data is abundant, Techies find ways to make it both valuable and fun appeared first on Mobility Lab.