Python Machine Learning Conference & GeoPython 2020

»Hurricane Damage Estimation using Siamese Networks«
2020-09-22, 15:55–16:15, Room 2

The aim of the project is to quantify the damage done to properties by hurricanes in a swift, automated and scalable way with the help of satellite imagery

The project takes in both the pre and post-disaster imagery of the affected area and tries to observe the change over time. A Siamese network is trained to observe the building changes between both the images, assigning each building a damage score from 0 (unaffected) to 3 (completely destroyed). These damage scores are fundamentally crucial for insurance companies for disbursing the correct loan amounts to the ones who've suffered the most.