Oliver Roick

Similarity Search

Explore satellite-image archives by finding similar images.

2022 - 2023

Client

NASA Earth Data

Tech

  • React
  • Chakra UI
  • Tanstack Query
  • Mapbox GL
  • Turf

Links

Using self-supervised learning, Similalaity Search allows researchers to select a sample image and to find similar images from am archive of satellite data. Similarity Search’s goal is to support scholars in climate-change research to find relevant data about wildfires, oli spills or hurricanes without manually combing through vast troves of archives often spanning several decades.

Work delivered

  • Major refactoring of the application, separating application pages and shared components, increasing re-usability and code efficiency.
  • New features, including:
    • Download results as CSV, the download being generate on the fly in the front-end
    • Improved results display, including image preview on the map and table view
    • Geocoding and reverse geo-coding