Dave Luo from Anthropocene Labs gives a beginnerfriendly talk that walks through a common MLformapping workflow (mapping building footprints from overhead imagery) with emphasis on what/why/how labeled data is important, defining key geospatial ML terms and concepts, and how training datasets are enhanced by using COG and STAC. As an example, he also introduces the Open Cities AI Challenge dataset of drone imagery (as COGs) and building footprint labels across 10+ African cities organized in a STAC.