The SpaceNet 8 Challenge, Flood Detection Using Multiclass Segmentation

Register
Submit a solution
The challenge is finished.

Challenge Overview

Natural disasters such as hurricanes, tornadoes, earthquakes, and floods inflict significant damage on infrastructure each year costing lives, time, and billions of dollars. There is a need to rapidly create maps and assess damage to assess the scale of damage and direct first responders. Utilizing the datasets and winning algorithms of previous SpaceNet challenges applied to a real- world disaster response scenario we hope to find answers to these questions: ��� How has building and road extraction improved from prior challenges? ��� What is the impact on performance for a multi class feature extraction challenge? ��� How accurately can roads obstructed from flood waters be characterized from pre- event roads and post- event satellite imagery? from satellite imagery? ��� Do flood assessment algorithms generalize to other geographic areas? TODO Please read above