Open AI Caribbean Challenge: Mapping Disaster Risk from Aerial Imagery
October 8th, 2019
WeRobotics and partners have carried out a number of Open AI Challenges over the years to help translate aerial data into actionable analytics for decision making. We have held these Open AI Challenges in the South Pacific, for example, and in Tanzania most recently. We now turn our attention to the Caribbean. Communities living in housing that is not up to appropriate construction standards are at the highest risk from a natural disaster. This risk is especially real for people living in poverty and informal settlements. While buildings can be modified to withstand disasters better, identifying those buildings is time-consuming, labor-intensive, and exorbitantly expensive.
That's why we're teaming up with the World Bank Global Program for Resilient Housing to help governments across the Caribbean use drone imagery to more quickly and more cost-effectively identify buildings that need to be prioritized for fortification against disasters. Drone imagery and AI can help accomplish this by automatically identifying the roof construction material, which is an important risk factor for earthquakes and hurricanes, and a useful predictor for other risk factors, such as building material.
In this $10,000 challenge, applicants will use high resolution aerial imagery from St. Lucia, Guatemala and Colombia to help develop machine learning classifiers that can automatically detect roof materials. A robust machine learning model that can accurately map the disaster risk for these areas will help identify priority buildings more quickly while driving overall costs down, thus increasing the impact of this resiliency effort.
Many thanks to all our partners, including the World Bank, DrivenData and MathWorks for making this challenge possible. MathWorks, makers of MATLAB and Simulink software, is sponsoring this challenge along with the prize for the top MATLAB user. They're also supporting participants by providing free software licenses and learning resources. To find out more about the challenge and to participate, please visit DrivenData’s Challenge website.