Menu

spacenet.ai

Accelerating Geospatial Machine Learning

Kelly Schroeder

Cross-Modal Satellite Imagery Registration

By

SN9: Cross-Modal Satellite Imagery Registration

The Problem

Swift and effective disaster response often relies on the integration and analysis of diverse remote sensing data sources such as electro-optical and Synthetic Aperture Radar (SAR). However, the co-registration of optical and SAR imagery remains a major challenge due to the inherent differences in their acquisition methods and data characteristics. SpaceNet 9 aimed to address this issue by focusing on cross-modal image registration, a critical preprocessing step for disaster analysis and recovery.
Participants in this challenge developed algorithms to compute pixel-wise spatial transformations between optical imagery and SAR imagery, specifically in earthquake-affected regions. These algorithms were evaluated for their ability to align tie-points across modalities, enabling better downstream analytics such as damage assessment and change detection.
To support the competition, the challenge provided a dataset consisting of high-resolution optical imagery from the Maxar Open Data Program and SAR imagery from UMBRA. The dataset includes manually labeled tie-points to evaluate registration quality. A baseline algorithm was also provided to set a performance benchmark, but participants are encouraged to explore novel approaches to significantly improve accuracy.
The objective of this challenge was to create algorithms that take two input images—an optical image and a SAR image—and output a two-channel transformation map. Each channel of the output image represents the shifts in the x and y directions required to align the optical image with the SAR image.
The accuracy of the output was evaluated using tie-points, which were manually identified matching features (e.g., road intersections) between the two images. The spatial transformation predicted by the algorithm was then applied to the tie-points in the optical image, and the alignment quality was assessed based on the distance between the transformed points and their corresponding reference points in the SAR image.
Further information on the SpaceNet 9 Challenge, dataset, baseline algorithm, and scoring metrics are available here.
Training data
https://spacenet-dataset.s3.us-east-1.amazonaws.com/spacenet/SN9_cross-modal/train.zip
Testing data
https://spacenet-dataset.s3.us-east-1.amazonaws.com/spacenet/SN9_cross-modal/testpublic.zip

RELATED BLOGS

License

The SpaceNet Dataset by SpaceNet Partners is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

Flood Vulnerability Detection Challenge Using Multiclass Segmentation

By

SN8: Flood Detection Challenge Using Multiclass Segmentation 

The Problem

Each year, natural disasters such as hurricanes, tornadoes, earthquakes and floods significantly damage infrastructure and result in loss of life, property and billions of dollars. As these events become more frequent and severe, there is an increasing need to rapidly develop maps and analyze the scale of destruction to better direct resources and first responders. 

To help address this need, the SpaceNet 8 Flood Detection Challenge will focus on infrastructure and flood mapping related to hurricanes and heavy rains that cause route obstructions and significant damage. The goal of SpaceNet 8 is to leverage the existing repository of datasets and algorithms from SpaceNet Challenges 1-7 and apply them to a real-world disaster response scenario, expanding to multiclass feature extraction and characterization. 

Since its launch in 2016, SpaceNet has made significant progress advancing open-source building footprint and road extraction algorithms. During SpaceNet 8, challenge participants will train algorithms on imagery and labels from previous challenges—as well as newly created labeled training datasets from Maxar—to rapidly map an area affected by flooding. Any winning open-source algorithm from SpaceNet 1-7 may also be used. New areas of interest (AOIs) will include New Orleans, Louisiana, following Hurricane Ida in August 2021; Dernau, Germany, during the June 2021 floods across Western Europe; and a new “mystery city” for blind testing of the algorithms. 

SpaceNet 8 aims to answer these questions:

  • How have algorithms that extract buildings and roads improved since SpaceNet was launched, and how can top algorithms from previous challenges be leveraged? 
  • What is the impact on performance for a multiclass feature extraction challenge—i.e., buildings and roads? 
  • How accurately can roads obstructed by flood waters be characterized from pre-event road detections and post-event satellite imagery? 

RELATED BLOGS

Training Data

aws s3 cp  s3://spacenet-dataset/spacenet/SN8_floods/tarballs/Germany_Training_Public.tar.gz . 
aws s3 cp s3://spacenet-dataset/spacenet/SN8_floods/tarballs/Louisiana-East_Training_Public.tar.gz .

Testing Data

aws s3 cp  s3://spacenet-dataset/spacenet/SN8_floods/tarballs/Louisiana-West_Test_Public.tar.gz .

Catalog

  • The data is hosted on AWS as a Public Dataset. It is free to download. An AWS account is not required; however, you must have the AWS CLI installed to access the data.  
  • To explore the dataset: 
    aws s3 ls s3://spacenet-dataset/spacenet/SN8_floods/
Register for SpaceNet 8

SpaceNet 8 – The Detection of Flooded Roads and Buildings publication 

License

The SpaceNet Dataset by SpaceNet Partners is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.