This document discusses improving open land use maps using satellite imagery. It aims to develop an algorithm to regularly update open land use maps based on satellite imagery like Sentinel 2 data. Currently, large areas of Africa and other regions lack detailed land use data. The algorithm will use machine learning and convolutional neural networks trained on labeled samples from Sentinel 2 imagery and open land use maps to classify land use in new satellite images. The goals are to collect training samples from Sentinel 2 and OpenStreetMap data and train and validate an initial classification model.