Or rsvp by reply! That works too.
2016-10-31 16:14:41
6
WEDNESDAY @CMU: Computer Vision Data Collection for Safer Biking and Walking
Pretty cool talk on Wednesday at lunch here at CMU. If you can't make it in person, check the livestream link and watch it from your desk or once it's posted to the Traffic21/T-SET website. If you can make it email me at nndimbie at andrew.cmu.edu so make sure we have enough (free) foods.
Traffic21 Faculty Seminar with Bernardo Pires: Computer Vision Data Collection for Safer Cyclists and Pedestrians
Wednesday, November 2
12:00 - 1:00PM
Hamburg Hall 1002
Lunch provided
This talk will focus on the use computer vision to collect actionable data for government officials and advocates that promote bicycling and walking. Although the health and environmental benefits of a non-automobile commute are well known, it is still difficult to understand how to get more people to take up active transportation. Infrastructure can have a dramatic effect on cycling and walking adoption, but represents a significant outlay of government resources. Thus, concrete usage and safety statistics are paramount for assessing and optimizing such spending. The talk will discuss three projects that are in various stages of development and use computer vision to collect actionable data: 1) A cyclist and pedestrian counting system for bike lanes; 2) A pedestrian detection and wait time measurement system for crosswalks; and 3) A speedgun app for mobile devices.
Bio:
Bernardo Pires is a Project Scientist at the Robotics Institute, Carnegie Mellon University. His recent work focuses on practical applications of Computer Vision, including wearable and hand-held devices, intelligent bike and pedestrian counting, and gps-denied vehicle navigation. From 2011 to 2014, he was a Post-Doctoral Fellow working on methods for gaze tracking and its applications to people and object recognition. His post-doctoral research technology was incorporated into a CMU spin-off. Bernardo holds a PhD degree from Carnegie Mellon University and a Licenciatura (B.S.) from Instituto Superior Técnico (IST), at the Technical University of Lisbon, Portugal. At IST, his research won the Prof. Luís Vidigal Award for the Best Graduation Thesis in the fields of Electrical Engineering and Computer Science.
Traffic21 is a multi-disciplinary research effort of Carnegie Mellon University. Its goal is to design, test, deploy and evaluate information and communications technology based solutions to address the problems facing the transportation system of the Pittsburgh region and the nation. Learn more here.
Can't make it? Catch the livestream on Wednesday (starting at ~12:10pm) here: https://mediaservices.cmu.edu/media/Traffic21+Live+Webstream/1_3hrrl27r
2016-10-31 16:13:51
BUMP! Don't forget that this is today.
If you're watching on the livestream and want me to ask a question for ya, reply to this thread.
xoxo
2016-11-02 10:01:50
I went to this talk by Bernardo Pires. It was quite interesting. I'll summarize what I learned:
Broadly, their goals are the use of computers and cameras to help city planning and transportation. Bernardo talked about 3 projects:
- Use of a portable camera on a pole that they can set up easily (with permission) by fastening it to an existing light pole. Fixed camera 10 or 15 feet in the air shoots video for hours, which they can use to count pedestrians or cyclists, for example. They used one near Phipps to count pedestrians and cyclists on sidewalk and in bike lane. They achieved 95% accuracy.
- Studying the effect of adaptive traffic signals (like the Surtrac hardware & software that's in use at some East Liberty intersections) on pedestrians. They did an experiment observing the Aiken&Centre intersection using computer vision techniques similar to project (1), calculating pedestrian waiting times at the traffic signal, both when Surtrac was turned on at that intersection and when it was off, to test the hypothesis (reported by some) that Surtrac (though good for cars) was bad for pedestrians. In their four days of testing, they found some evidence for that, though the effect was small: average pedestrian wait time before crossing rose from 17 seconds without Surtrac to 19 seconds with Surtrac, on the days they observed.
- They're developing a smartphone app to act like a radar gun to measure the speed of cars. Currently, the user must hold the camera still, record video of the car driving past, with license plate visible, use the app to point out the corners of the license plate in two frames of video. Knowing the size of a license plate and the time difference between the frames and some camera characteristics (e.g. camera lens focal length) the app can estimate car speed. They're working on improving their software e.g. to make it work when the camera is not held still, and they hope it will empower citizens to measure traffic speeds without expensive equipment like radar guns.
2016-11-04 23:14:57
How often do police nationwide replace their speed guns? Seems they should be functional enough for a suitably informed citizen to operate, accurate enough for our purposes, and if gotten at scrap value, should be affordable, easily obtained from someone who deals in that sort of stuff.
Granted, a citizens group isn't going to write tickets, but they could certainly collect data to assist with identifying problem areas, or the extent to which speeding is a problem in a given area.
2016-11-05 20:54:02
Great summary, Paul. Thanks. Video not available just yet. I'll post it as soon as they send it to me.
Stu that's a novel idea. I think the merit of Pires' app, as he's aware, is that few Pittsburghers walk around with radar guns, but hundreds of thousands walk around with smartphones.
Are you both going to Transportation Camp on the 19th? http://transportationcamp.org/events/pgh-2016/
2016-11-06 11:05:03