João Caldeira, Alex Fout, Aniket Kesari, Raesetje Sefala
UPDATE: We are pleased to announce that this job group won a Highlighted Paper Award at the AI For Social Good NIPS2018 Workshop! Congratulations to the Jakarta Fellows!
Improving Traffic Safety in Jakarta Through Video Analysis
Fig. 1: A satellite map of Jakarta with traffic cams marked.
The World Health Organization (WHO) approximates that over 1.25 million individuals pass away each year in traffic mishaps. Nearly 2000 such deaths happen each year in Jakarta, Indonesia alone, making it among the most hazardous cities worldwide for traffic safety. These deaths are terrible, but a lot of them are preventable through effective city preparation. This summertime, our team at the Data Science for Social Good Fellowship (DSSG) at the University of Chicago set out to help the city of Jakarta bring this figure down and safeguard its residents from traffic-related injury and death.
Figure 2: One may envision hiring a professional to view a video and transcribe occasions of interest.
Device knowing tools supply the methods to utilize traffic video to their maximum prospective mainly by permitting the processing of video at scale. One could imagine employing a technician to continually view a video feed of some specific crossway, and count the number and kinds of cars they see, along with fascinating events such as a vehicle driving down the incorrect side of the road. This technique would be cumbersome even for one video feed, but ends up being abstruse when scaling to hundreds or countless cams. It just would not deserve the human resources included to attempt to glean details from raw videos in a manual style.
Jakarta is a city in transition, having actually experienced explosive population development over the last couple of decades. These factors undoubtedly lead to traffic events. Our objective was to allow the city to lower the number of these events through city planning.
From a data science viewpoint though, one of the core obstacles in attending to traffic security problems is that it is hard to gather high quality traffic information. The city also has a promising resource: thousands of traffic video cameras posted throughout the city, with many more to come.
Figure 3: The issue is, having human watch videos does not scale well at all!
This is a vehicle …
Over the summer, we gained a strong understanding of why traffic safety is a challenging issue to resolve, what the core difficulties are in addressing it, and how information science may assist. Moving forward, we hope these insights will lead the way for cities to get rid of traffic security threats altogether. We hope that our work moves us closer to a world where traffic accidents are not considered an inevitability, and we give individuals the security of knowing that they can safely move about their cities for work and play.
Almost 2000 such fatalities take place yearly in Jakarta, Indonesia alone, making it one of the most hazardous cities in the world for traffic security. From an information science perspective though, one of the core challenges in attending to traffic safety issues is that it is challenging to collect high quality traffic data. The city also has a promising resource: thousands of traffic cameras published throughout the city, with many more to come.
We saw this issue as a chance to showcase the potential for information science by building a system where computers do the work of processing video, freeing humans to do analysis and planning. We looked for to create a pipeline that might take in raw video and output a structured database. This database would consist of information about activity on the highways, and hence assist in policymakers work by providing them with rich info about car behaviors throughout the city.
Understand the “context” of a scene (i.e. what is a roadway, a sidewalk etc.).
Together with our partners, we focused on several fascinating and useful top priorities. In the medium term, we found out that Jakarta had an interest in better management of its traffic signals and implementation of “traffic stewards.” More effective interventions in these two areas would assist the city handle self-important blockage, and perhaps prevent accidents developing from bad traffic circulation management. In the long term, the city was interested in understanding what kinds of facilities modifications might enhance security. Setting up a mean may avoid automobiles from crossing into oncoming traffic, or developing a bus lane might enhance congestion along a popular path.
Rather, we spent much of our time comprehending the criteria of the social issue. As soon as we understood this aspect, we were able to map technical solutions to policy interventions.
… that is moving …
Because traffic safety is such a broad topic, much of our work was driven by the ultimate interventions that our partners hoped to inform. From our point of view, numerous technical choices might eventually just be made after we comprehended the policy interventions that particular methods were meant to facilitate. We therefore sought to understand the medium- and long-lasting interventions that our partners wanted, so that we might focus our deal with gathering the information that would help direct those interventions most effectively.
Developing this pipeline was not simply a technical difficulty. Effective advancement depended on a deep understanding of the social context in Jakarta. We might not begin making decisions about pipeline style, computer vision methods, and recognition without understanding our partners top priorities, and their meaning of “traffic security.” Behaviors such as having more than two individuals on a bike, bring a food cart through a street, and weaving through traffic were surprising to some members of our team, and we needed to learn the cultural context that specified to Jakarta. This was a crucial step because importing our own prejudgments and predispositions about traffic patterns in the U.S. and other parts of the world would have ultimately not served Jakarta well. Through a number of discussions with our partners at Jakarta Smart City and Pulse Lab Jakarta, we gradually defined the scope of the problem that we were facing.
Figure 4: Computers are much better at handling tasks at scale, therefore maximizing people to do the effort of analysis and planning.
We picked these objectives because accomplishing these would enable the interventions that Jakarta was eventually thinking about deploying. Think about, for example, that a person of the primary problems that our partners were worried about was lorries driving on the wrong side of the roadway, and for that reason developing an immediate safety threat. For a computer system to successfully record such an occasion, it needs to understand that there is an automobile, understand that it is moving in a particular direction, and know what the “incorrect” instructions is on a street. Alternatively, another issue might be scooters and motorcycles driving on pathways and threatening pedestrians. Once again, the computer would need to comprehend what a motorcycle is, know that it is moving and not parked, and recognize that it is taking a trip on a pathway and not a roadway.
Figure 5: Our task was to take raw video footage (pictured top), and extract details like what type of lorry is being spotted, what direction is it relocating, and what surface area is it taking a trip on.
Price quote the instructions of motion for objects throughout frames.
As soon as we comprehended the scope of interventions, we were then able to develop a prepare for the data science techniques that might best address the general problem of improving traffic safety. In particular, we determined 3 central objectives:.
We then turned to pulling together the best techniques for each objective into a cohesive pipeline as soon as we defined goals. Integrating this information permitted us to produce a powerful tool that we hope will help Jakarta in its efforts to improve traffic safety throughout the city. By the end of the summer season, our pipeline had the ability to effectively identify when an automobile was carrying on the incorrect side of the road, and flag this event in the database. We anticipate that in the future, Jakarta will have the ability to recognize a variety of occasions of interest, and construct a database which contains rich info about traffic habits throughout the city.
Figure 6: To show the power of our pipeline, notice how in this one crossway we see multiple examples of lorries taking a trip on the incorrect side of the roadway. Our pipeline automatically flagged these occasions within the very same three-day span, and in fact all of these other than the leading left took place within the same 2 hours. Policymakers will have the ability to use the knowledge that individuals tend to drive on the incorrect side of the roadway at this specific time to much better understand when to deploy traffic stewards, or what kinds of facilities to develop.
We want our work this summertime to offer a plan for how cities around the globe might approach the job of establishing effective wise city efforts. Traffic cameras are wonderful examples of the sensors that numerous cities have explore in the previous few years. More than at any time in history, cities have an unprecedented capacity to gather top quality details about how people move in a metropolitan landscape. Our code and technical methods need to highlight the potential for utilizing such sensors in 21st-century urban planning.
Integrating this information enabled us to develop a powerful tool that we hope will aid Jakarta in its efforts to improve traffic security throughout the city. We expect that in the future, Jakarta will be able to identify a variety of events of interest, and develop a database that includes abundant info about traffic habits throughout the city.
… on the road.
Find and classify things in a frame.
Raw Video Footage.
