SpaceNet 8 - The Detection of Flooded Roads and Buildings

Image | Flood | Preparedness

SpaceNet 8 is flood-disaster dataset for building detection, road network extraction and flood detection. It covers 850km^2, including 32k buildings and 1,300km roads. It is introduced for multi-class segmentation and binary classification.

  • ML task type: Multi-class segmentation, binary classification
  • Data Source: Earth Observation Data and GeoSpatial Imagery (Satellite)
  • Size: Images cover 850km^2, 32k buildings, 1,300km roads
  • Timespan: July 2021, August 2021, and other not specified
  • Geographical Coverage: Germany, Louisiana, mystery location
  • Baseline Information
  • Evaluated on: U-Net with ResNet34 encoder for segmentation; Siamese CNN for flood predictions
  • Metrics used: Intersection over Union (for building flood assessment), Average Path Length Similarity (for road network prediction)
  • Results as reported in original paper: IOU: 0.66; APLS: 0.45

Ronny H¨ansch, Jacob Arndt, Dalton Lunga, Matthew Gibb, Tyler Pedelose, Arnold Boedihardjo, Desiree Petrie, and Todd M Bacastow. Spacenet 8-the detection of flooded roads and buildings. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 1472–1480, 2022.