Annotating Imagery at Scale with GeoVisual Search

How to Go From Zero Data to Thousands of Annotated Satellite Images

The GeoVisual Search tool in action. In the main window, a user clicks an area of interest highlighted by a red box. Once clicked, on the top-left, the navigation map populates with red boxes where visually similar features were found and allows for click-navigation around the map. Below the navigation map, the top 1000 results from the search are shown and each can be clicked to zoom to that particular feature.

Introduction

Global data discovery is hard and time-consuming

Descartes Labs GeoVisual Search

Each tile from a GVS imagery composite is fed through the GVS encoder which generates a vector with 512 binary values. In this case, 5 images generate 5 vectors represented here as a matrix.
Example tiles from NAIP (left) and Landsat (right) where each image is centered on the same latitude and longitude and are 128 x 128 pixels in size. Notice the huge difference in the land area covered by each.

Center Pivots Example

Some of the center-pivot templates used to generate a large machine learning training set.
After a user clicks a feature, the URL changes to include the dlkey which uniquely identified that box on the Earth.

Data Annotation

Conclusion

References

--

--

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store