A supervised Random Forest land cover classification in QGIS, using the dzetsaka plugin as a no-code front-end to scikit-learn. Over seven steps it covers acquiring a Sentinel-2 L2A scene of Masbate from the Copernicus Data Space, resampling and stacking the ten usable reflectance bands, installing dzetsaka and scikit-learn, digitising training polygons for six land cover classes, running the train/predict algorithms, cleaning speckle with Sieve and a majority filter, and assessing accuracy.
This course introduces plain-language prompts with the GeoAgent plugin in QGIS. It covers install, LLM setup (Ollama recommended: local, free, no data leaves the room).
Under development.