Next-Gen: Diving Into Deep Learning for Social Good

DrivenData is an organization that has hosted online challenges since 2014, allowing people from around the world to come together to solve social challenges. The platform’s competitions, like its Open AI Caribbean Challenge, combine the latest technologies in data science and crowdsourcing to help solve critical problems in areas such as health care, education and public services. 

In partnership with companies such as MathWorks, the Open AI Challenge focused on better preparation for natural disasters in the Caribbean. The competition hosted 1,425 participants, including engineers, scientists and students, from over 150 countries. These competitors contributed to the 2,753 entries aimed at developing the best algorithms for social good. A discussion panel accompanied the competition, where participants taught and learned from others in the data science community. 

Neha Goel is a deep learning technical evangelist for student competitions for MathWorks. We spoke to Goel to gain more understanding of this competition. Here’s how the conversation went.

Digital Engineering: Can you tell us about some of the designs that are part of the event and how they came to be?

Neha Goel: DrivenData’s Open AI Challenge tasked competitors with developing AI [artificial intelligence] models that can identify areas in the Caribbean most vulnerable to natural disasters. The aerial drone imagery data for the challenge was provided and prepared by WeRobotics, a non-governmental organization dedicated to bringing robotics technology to developing countries, and the World Bank Global Program for Resilient Housing.

The competition gave participants the opportunity to work with this unique aerial imagery. The AI models developed by competitors needed to analyze the aerial drone imagery data and identify high-risk areas based on roofing materials—a critical determining factor for the degree of damage a structure might suffer during an earthquake or flood. 

DE: Can you provide some examples of what the event has produced or what you expect it to produce?

Goel: The competitors’ algorithms can help improve disaster resilience efforts by prioritizing areas or specific buildings for greater protection in advance of disasters. This approach will save tremendous time and risk management resources, both of which are in short supply when natural disasters hit. 

In conversations with Greg Lipstein, cofounder and principal at DrivenData, he discussed how identifying at-risk areas and buildings previously required personnel to go door to door on foot, a taxing process from a budgetary and time perspective. With citizens’ safety on the line, the approaches outlined in the Open AI Challenge help mitigate these time-sensitive obstacles. The competition’s winning solutions are all released under an open-source license for anyone to use and learn from. 

Along with benefitting efforts in the Caribbean, DrivenData’s challenge also gave engineers, scientists and students the opportunity to work with a unique, real-world dataset as provided by WeRobotics and the World Bank and develop AI models with life-saving potential. This allowed the competitors to gain invaluable experience working with real-world data, and the competition platform enabled participants to engage and connect with thousands of individuals around the world who are passionate and motivated to use data science for social good.

I’ve also been able to connect with competitors such as Ning Xuan, a biomedical engineer with a master’s degree from Columbia University and the MATLAB bonus award winner, who entered the competition to enhance his deep learning skills. The competition gave him hands-on experience to work on a project with real-life data and a broader perspective on how data science can be applied to multiple fields of work. One of his key learnings was understanding that geolocation coordinates are mapped on a 3D surface to simulate the Earth’s geography, which he didn’t know prior to the competition.

DE: Does MathWorks have a particular stance on adopting an innovation that is linked to the program? What drove MathWorks to sponsor the event and coordinate it?

Goel: MathWorks provided competitors access to MATLAB for the competition as part of its sponsorship. Participants used MATLAB and Deep Learning Toolbox to process the geographic and aerial imagery data of buildings in St. Lucia, Guatemala and Colombia; classify images of roofing materials; and develop and train the deep learning model. 

Today, machine learning and deep learning are used to solve these pressing dilemmas facing humanity, and MathWorks acknowledged this through sponsorship of the competition and by offering MATLAB user awards to successful competitors. 

For example, Ning’s model correctly identified the roof construction material with over 80% accuracy, an impressive result, given five different roof categories to choose from. This was more than double the accuracy rate of the pre-competition benchmark model. 

Companies, organizations and governments have always looked to use technology to solve the world’s problems, and MathWorks is no exception. As engineers and scientists continue to be called upon to apply their expertise to deliver social good, competitions like DrivenData’s help foster AI skills and experience working with real-world data.

More MathWorks Coverage