I am a Data scientist on Microsoft's AI for Good team.
PEARL, the Planetary Computer Land Cover Mapping Platform, uses state of the art ML and AI technologies to drastically reduce the time required to produce an accurate land cover map. Scientists and mappers get access to pre-trained, high performing starter models, and high resolution imagery (e.g. NAIP) hosted on Microsoft Azure. The land cover mapping tool manages the infrastructure for running inference at scale, visualizes model predictions, shows the model’s per class performance, allows for adding new training classes, and allows users to retrain the model. The tool helps harness the power of expert human mappers and scientists through immediate model evaluation metrics, easy retraining cycles and instant user feedback within the UI.