Deep Learning Based Damage Detection on Post-Hurricane Satellite Imagery

This repo contains the materials presented in the paper "Deep Learning Based Damage Detection on Post-Hurricane Satellite Imagery" submitted to the INFORMS 2018 Best Student Paper Competition.

The repo contains the following items:

• Source python code for image classification
• Source python code for extract, filter, crop the bounding boxes from raw satellite imagery
• Source code to download raw satellite imagery from Digital Globe website
• Source code to manipulate geoTIFF files using gdal library
• The short paper in PDF
• The poster presentation in PDF
• The dataset:
  1. train_another : the training data; 5000 images of each class
  2. validation_another: the validation data; 1000 images of each class
  3. test_another : the unbalanced test data; 8000/1000 images of damaged/undamaged classes
  4. test : the balanced test data; 1000 images of each class