Mapping 57 million agricultural fields with AI

Oct 30, 2018

Powered by an impressive combination of AI, machine learning, computer vision, and data visualization, OneSoil maps 57 million agricultural fields across 46 European countries and the US.

The platform, available for free on web and mobile, is changing the way farmers interact with their fields and crops by digitizing their workflow. The map allows users to explore individual, delineated fields by crop type and production score, and compare changes over the last three years. This information can be used to help farmers avoid sketching fields by hand, traders to accurately estimate global harvest, insurers to accurately assess their risks, and it gives communities the resources to work towards global food security.

Users can seamlessly toggle through 4 different types of map layers, 20 different crop types, and 3 years of data.

OneSoil takes publicly available satellite imagery from Copernicus, the EU’s Earth Observation program, and compresses it to allow processing in real-time. The imagery is enormous — OneSoil compresses over 20 petabytes of data down to 100 terabytes, then uses AI to automatically clean the imagery, delineate field boundaries with a spacial resolution of five meters, and determine crop type based on 20 of the world’s most popular crops. The platform is automatically updated hourly.

OneSoil uses vector tiles to keep the map lightweight, fast, and interactive, without sacrificing on map performance. Mapbox GL is essential to rendering the map at 60 frames-per-second, while styling on the fly. Users can take advantage of custom layers to add 3D and advanced data visualization with just one line of code, use data-joins to create a live updating map, and easily integrate hover, highlight, and filter effects for users clicking, touching, or mouse’ing-over data.

Check out everything that’s possible with our newest version of Mapbox GL. Still using raster tiles? Take advantage of our transition guide which makes the switch to vector tiles painless.