(Image source)

Automatic Damage Annotation on Post-Hurricane Satellite Imagery

The Disaster Damage Detection team worked on one of three projects from the 2018 Data Science for Social Good summer fellowship at the University of Washington eScience Institute. The goal of this project is to use post hurricane satellite imagery data to train object detection models to automatically detect damages from satellite images after hurricanes to facilitate the damage assessment process for emergency managers.

This website catalogs the groups work to:

  1. Gather, clean and prepare training datasets with annotations for damaged buildings and test datasets

  2. Train the machine learning algorithm using the training data

  3. Test the model on a test dataset

The Team

Project Lead: Youngjun Choe, Assistant Professor of Industrial & Systems Engineering and Director of the Disaster Data Science Lab, Aerospace & Engineering Research Building, University of Washington

Data Scientist Lead: Valentina Staneva, Senior Data Scientist, eScience Institute

DSSG Fellows: Sean Chen, Andrew Escay, Chris Haberland, Tessa Schneider, An Yan