Data Science for Social Good Data Fest 2018 - Jakarta

Summary

This video discusses the traffic safety challenges in Jakarta, Indonesia, where traffic accidents claim thousands of lives each year. The city faces significant risks due to congestion and hazardous events, such as pedestrians walking on roads and vehicles driving in the wrong direction. The solution presented involves using machine learning to analyze footage from over 5,000 traffic cameras to identify, track, and classify incidents automatically. The technology helps city planners make data-driven decisions to improve road safety, infrastructure, and public transport. This system could be transformative not only for Jakarta but for cities globally facing similar issues.

Takeaways

• 😀 Traffic-related deaths are a significant issue, with 1.5 million lives lost annually worldwide, including 2,000 in Jakarta, Indonesia.
• 😀 These deaths are largely preventable with smart city planning and data-driven interventions.
• 😀 Jakarta, with a population exceeding 10 million, faces severe congestion and traffic accidents due to rapid urban growth and outdated infrastructure.
• 😀 The city has identified various events that increase traffic safety risks, such as street vendors blocking sidewalks and wrong-way driving.
• 😀 Jakarta plans to improve traffic signal management and deploy traffic stewards in the medium term, and inform infrastructure planning in the long term.
• 😀 A key challenge is that Jakarta lacks the means to effectively monitor where and how traffic safety risks occur, despite having over 5,000 traffic cameras.
• 😀 Manually analyzing footage from 5,000 traffic cameras would require 44 million person-hours annually, making the process inefficient.
• 😀 Machine learning can help by automating the extraction of useful information from traffic camera footage, allowing human resources to focus on problem-solving.
• 😀 The team developed an end-to-end system that identifies pedestrians, bicycles, motor vehicles, tracks their movements, and classifies their location on roads or sidewalks.
• 😀 The system can create detailed event narratives, such as detecting cars traveling the wrong way and avoiding traffic jams, which helps inform city planning decisions.
• 😀 The system's data could be used to make informed decisions about traffic stewards' placement or where new crosswalks should be built, significantly enhancing urban planning and safety.

Q & A

What is the primary cause of traffic-related deaths in Jakarta?

The primary cause of traffic-related deaths in Jakarta is increased congestion and traffic accidents, which are exacerbated by the city's rapid growth and underdeveloped transportation infrastructure.

How many people die annually in traffic-related incidents worldwide, according to the World Health Organization?

The World Health Organization estimates that 1.5 million people die in traffic-related incidents every year.

How many traffic cameras does Jakarta have to monitor the roads?

Jakarta has over 5,000 traffic cameras monitoring the city's roads.

What challenges does Jakarta face regarding its traffic safety efforts?

Jakarta faces the challenge of managing and analyzing the large amount of data generated by its traffic cameras. The city lacks a practical means of extracting useful insights from this data.

How much human effort would be required to monitor the traffic cameras manually?

Manually monitoring all 5,000 traffic cameras would require approximately 44 million person-hours every year, which is highly impractical.

What role does machine learning play in addressing Jakarta's traffic safety problems?

Machine learning is used to automate the process of extracting useful information from traffic camera footage, reducing the need for human labor and allowing more time for decision-making based on the data.

What types of objects can the developed system identify from traffic camera footage?

The system can identify pedestrians, bicycles, and motor vehicles, as well as track their position, direction of movement, and whether they are on the road or sidewalk.

What kind of incidents is the system designed to identify?

The system is designed to identify hazardous events, such as vehicles driving the wrong way or pedestrians and cyclists in dangerous locations, such as on the road instead of the sidewalk.

How did the system perform in detecting traffic violations over a 3-day period?

Over a span of three days, the system identified four incidents where cars left their lane to bypass traffic jams, with three of these events occurring within two hours.

How could city planners use the data collected by the system?

City planners could use the data to inform decisions about where to place traffic stewards, install crosswalks, and improve road infrastructure based on real-time or past data on high-risk incidents.