Dynamic World, Near real-time global 10 m land use land cover mapping
Abstract
We show a new automated approach for globally consistent, spatially explicit high resolution land use land cover (LULC) leveraging deep learning on 10m Sentinel-2 imagery. We also present a new open and global training dataset that can be used for the development of machine learning algorithms. Our method achieves state-of-the-art agreement with our validation data over other popular global and even regional LULC datasets, exceeding the next best agreement by 7.5% and 5.2% respectively. Together with Google Earth Engine and Google Cloud AI Platform, we utilize a highly scalable cloud based system for generating LULC maps and provide an open, continuous feed of LULC in parallel with Sentinel-2.
