geoai

Why we included this project

GeoAI ties the usual remote-sensing workflow into one Python package: you can search and download satellite imagery, build labeled image chips, train a PyTorch model, and run inference without juggling several separate tools. It leans on established libraries like PyTorch, Transformers, and segmentation models, so you get pretrained backbones rather than starting from scratch. Training and inference helpers handle the tasks that dominate earth-observation work, such as land-cover classification, object detection, and building footprint extraction, and the results can be written straight to GeoJSON or GeoPackage. That makes it a solid fit for researchers and GIS analysts who want a reproducible route from raw data to a finished map layer, without assembling a custom pipeline by hand.