The inspiration is to help improve disaster response in the age of climate change and frequent natural disasters.
What it does
The tool helps emergency responders and planners understand which areas are flooded and what is affected as a result. It pulls Sentinel-1 SAR satellite data from a public S3 bucket and runs it through a Convolutional Neural Network to produce a map of flooded areas. It then plots the inference results on a background of population, infrastructure and land-use/land-cover maps to show the extent of the impact.
How we built it
Satellite data and deep learning models were used to build the ML model that powers a map-based tool.
Challenges we ran into
Lack of adequate support. AWS office hours for EMEA region were literally a few hours to the deadline so teams from Africa did not get enough time and support for adequate submissions.
Accomplishments that we're proud of
Building a tool that will improve disaster response and save lives.
What we learned
How to leverage Sagemaker for machine learning projects.
What's next for Flood response tool
Improvements and refinements into a highly useful application.
Log in or sign up for Devpost to join the conversation.