Tutorial: Remote Sensing Image Classification with QGIS

For decades, digitising parcels in remote sensing and GIS workflows meant drawing polygons by hand. Anyone who has traced field boundaries or building footprints knows how time‑consuming this can be. Even with good basemaps, careful snapping, and plenty of patience, manual digitizing remains one of the slowest steps in many mapping projects.

In recent years, however, a new wave of tools has started to transform this process. QGIS now hosts an expanding ecosystem of plugins that use artificial intelligence to automatically segment objects in imagery. Many of these tools build on Meta’s Segment Anything Model (SAM), a foundation model trained on billions of image–mask pairs to identify and outline objects of almost any type, even in images it has never seen before. SAM doesn’t “know” what a parcel or a tree is; instead, it excels at finding coherent shapes and boundaries, making it a powerful engine for geospatial segmentation tasks.

In this chapter, we’ll work with one of the most accessible implementations of this technology: the AI Segmentation plugin by Terralab. It’s lightweight, easy to install, and designed to integrate smoothly into everyday QGIS workflows. With just a few clicks, you can generate clean parcel boundaries from high‑resolution imagery, dramatically reducing the time you spend digitizing.

Watch this video for a quick demo of the plugin:

Let's get started.