DigiFarm – detecting the world’s most accurate field boundaries to power precision agriculture

Initial submission

All precision agriculture services start with accurate field boundaries and seeded acres. Unfortunately, the problem is that critical decisions are made based on inaccurate data, cadastral field boundary data, which is affecting the entire agricultural value chain. DigiFarm has spent the last two years developing a deep-resolution algorithm for Sentinel-2 imagery to increase image resolution by 10x from 10m to 1m, coupled with a deep neural network model to automatically detect field boundaries on a large scale. DigiFarm has delinea- ted over 15 million ha across the world and has achieved an average (IoU) accuracy of above 0.96, which represents up to 20% greater accuracy than existing cadastral map data. DigiFarm delivers four key models to B2B and B2G clients through various API endpoints, including: automatic field boundary detection (and seeded acres), deep resolution of Sentinel-2 at 1m resolution on demand, automatic delineation of in-field productivity zones and crop classification, the ability to automatically detect field boundaries (seeded acres) using deep neural network models and super-high-resolution satellite imagery.