segment-geospatial

Why we included this project

Drawing boundaries on satellite and aerial imagery by hand is slow and inconsistent, so this package adapts Meta's Segment Anything Model to geospatial data to automate the work. It wraps SAM in a thin Python API, letting researchers and GIS analysts prompt for a feature and get vector results back in a few lines of code, with helpers for running inference over tiles, masking clouds, and exporting to common geospatial formats. That means you can pull out built-up areas, water, or vegetation from a scene without training a custom model every time. The convenience utilities also handle the coordinate and format wrangling that usually trips up remote sensing work, so outputs drop into existing raster and vector pipelines. If extracting shapes from imagery is a recurring task rather than a research project in itself, this is a solid shortcut.