The bad apple drivers affecting big apple commutes by robbing our lanes.
Alex Bell is a New York City based cyclist, computer scientist, and transit advocate whose “analytics as advocacy” has recently been featured in the New York Times.
ARGO facilitates a streets data collaborative that builds, operates, and maintains a low-cost data infrastructure that can be shared across cities to achieve better transit and infrastructure outcomes.
Every morning and evening, New Yorkers commuting by bus or bike are keenly aware of the pain and danger caused by cars and truck drivers who illegally park their vehicles in a designated bike or bus-only lane thereby robbing us of a right to a smooth commute.
Leading transit advocates, the Transit Center has invested heavily in its Bus Turnaround campaign that has shown how more than a 20% decline in bus ridership in the past decade and a half can contribute to severe congestion in the city. Their solutions appeal to the MTA and Transit Authority to significantly redesign core elements of the city’s bus system to serve New Yorker’s, a better bus ride.
Even the mega star Youtuber, Casey Neistat, has highlighted the menace of the Great New York Bike Lane Robbery (often committed by New York’s Finest).
These illegally-parked vehicles affect commute times and contribute to congestion by forcing buses to stop and merge into the regular traffic lane and bully bike riders into traffic, jeopardizing their safety and antagonizing drivers en-masse.
Every day, a majority of street commuters are collectively ripped off by a small minority of drivers trying to get ahead by a few seconds. Hasn’t this city heard that story before!?
Commuting death by a thousand lane blocks.
We can do better. Here’s how Alex Bell is a street data warrior for moving the needle on this issue.
Alex developed a machine learning algorithm, using publicly available, real-time image data from the New York City Department of Transportation, to study bus & bike-lane robbery in NYC.
The same traffic imagery that is broadcast every morning on your screens has been repurposed to measure what at first may seem small infractions but in the sum can be a major drag against those who use street-based public transit.
The results of Alex Bell’s analytics illustrate a critical gap in New York City’s traffic enforcement. They offer insights into how we can quickly, affordably, and equitably get public commutes back on track (lane) again.
To catch a Lane Predator.
Alex sought to examine this plain sighted problem of lane robbery using Ground Truth data. Something we here at ARGO are passionate advocates of, and share a common purpose with our Street Quality IDentification Project (SQUID) to measure street and bike lane quality using open source imagery.
After all in the words of a former 3-term Mayor of this city:
If you can’t measure it, you can’t manage it and you can’t fix it — Mike Bloomberg
Throughout the city, The Department of Transportation (NYC DOT) maintains several hundred traffic cameras that post images in real time to dotsignals.org.
Alex sourced images from December 2017 from a single camera on the corner of 145th Street and St. Nicholas Avenue in Upper Manhattan
St. Nicholas Avenue has unprotected bike lanes in both directions, features a bus stop, and is near a police station.
He then wrote a program that leveraged a popular machine learning framework developed by Google called Tensorflow. Alex trained his Tensorflow model on 2,000 labelled and pre-classified images of vehicles blocking the bike lane and unleashed it to measure lane robbery over 800,000 images of traffic.
The Results
On weekdays during a 12-hour window from 7 a.m. to 7 p.m, the bike lane in front of the traffic camera was blocked 57% of the time, and the bus stop was blocked 55% of the time.
Across the full ten-day period, 24 hours a day, the bike lane was blocked 40% of the time, and the bus stop was blocked 57% of the time.
Remember: This is just a single block’s snapshot of a problem that spans the 7,000+ center lane miles across all five boroughs of New York City. Bicyclists frequently have to swerve into traffic to get around vehicles. More often than not, buses have to park far from the curb, blocking traffic and forcing riders into traffic. Disabled bus riders can’t safely board, and the bus is slowed down as passengers have to thread between parked cars to get on and off the bus.
Next Steps
Alex’s study pointedly illustrates how a small number of bad apple drivers are capable of causing disproportionate delays to thousands of big apple commuters.
This study also offers a clear opportunity for the city to better enforce existing traffic laws, and reduce overall congestion through small gains in a thousand places across the entire city.
Keeping bus and bike lanes clear increase safety for drivers, bikers, pedestrians, and public transit riders alike.
The City could use this research improve traffic surveillance at high traffic intersections across the city. Improved imagery and blurring people’s faces using open source image processing and machine learning techniques can add a crucial layer of data to help reduce congestion through while preserving privacy.
The data can be of research value as well. Urban focussed data scientists and public technologists can use the open camera system to further study how lane blocking affects other aspects of congestion and traffic behavior.
Team ARGO is awe-struck by Alex’s contribution and thanks him for reaching out to us with his work. It is because of work like this that we are increasingly convinced and motivated that a growing movement of public technologists can design technology that stewards the public space by empowering cities to do more with less.
Takeaways and resources
- The source code of this entire project lives here.
- Tensorflow’s getting started guide proved to be particularly useful.
- Alex’s Intel Core i7–4770K took an entire day to retrain the last layer of the model. In total, 800,000 images were processed with the model to look for vehicle positions. Amazon Web Services was also used to analyze the imagery. A p2.xlarge with 1 Nvidia K80 GPU took approximately half a second per image. A large instance type could significantly improve speed.
- Determining a “blocking vehicle” was simply a matter of drawing a shape around a vehicle that was not a bus, and whether the center of this shape was in the lane.
- via Alex “I considered training the algorithm to try to spot bike lanes and bus stops and do its own determination. But I was concerned that with the street markings barely visible it would be very difficult. So instead, I just drew the lanes because I could visit the street and determine the actual markings. If this was to be used at scale there might have to be a manual portion of drawing or determining the lanes/stops and mapping that to pixels.”
