Training a Model in the Cloud Using Data from Radiant MLHub

Abstract:

The most challenging aspect of a machine learning pipeline is gathering enough high quality training data to build an accurate model. Radiant MLHub is an open repository for geospatial training data which can be accessed in a machine-readable fashion using a STAC-compliant API. Utilizing Radiant MLHub and other STAC-compliant APIs, researchers can programmatically mix and match training data and source imagery to speed up the process of training a model and increase the quality of its predictions. To enable a rapid iterative training process, the training can be run in the cloud with much more processing power and less upfront costs compared to a traditional desktop training process. In this presentation, I will review Radiant MLHub API and demonstrate a model training in the cloud using data hosted on Radiant MLHub.