Creating xBD: A dataset for assessing building damage from satellite imagery

Image | General | Response

xBD is a general-disaster dataset for change detection and damage assessment of buildings. It contains 850,736 building polygons across 22,068 images and 45,361.79 km^2. It is introduced for multiclass (ordinal) classification.

  • ML task type: Multiclass (ordinal) classification
  • Data Source: Earth Observation Data and GeoSpatial Imagery (Satellite)
  • Size: 850,736 building polygons across 22,068 images and 45,361.79 km^2
  • Timespan: 2011 - 2019
  • Geographical Coverage: Global
  • Baseline Information
  • Evaluated on: ResNet50 + shallow CNN
  • Metrics used: Weighted F1
  • Results as reported in original paper: Weighted F1: 0.2654

Ritwik Gupta, Richard Hosfelt, Sandra Sajeev, Nirav Patel, Bryce Goodman, Jigar Doshi, Eric Heim, Howie Choset, and Matthew Gaston. xBD: A dataset for assessing building damage from satellite imagery. arXiv preprint arXiv:1911.09296, 2019.