{"name": "AI Segmentation by TerraLab", "package_name": "AI_Segmentation", "description": "AI segmentation for QGIS: detect and extract building footprints, trees, vegetation or any object from your raster, then refine & export as vector polygons", "about": "Draw a zone on your imagery, type what you want (\"buildings\", \"trees\", \"solar panels\") and get every match back as vector polygons. Or click one object and get its outline. It replaces tracing by hand.\n\n\u2022 Automatic: every building, tree or other object inside the zone you draw, in one run, with a live preview. It runs on TerraLab servers, so no graphics card is needed\n\n\u2022 Semi-Auto: click an object, add positive and negative points, get a clean outline. Pick \"My computer\" and it runs locally and offline, or \"Cloud AI\" and nothing needs downloading\n\n\u2022 Works on the imagery you already have: orthophotos, drone and satellite images as GeoTIFF, JPG or PNG, and online basemaps (WMS, XYZ tiles)\n\n\u2022 Review before you keep: right angles for buildings, expand, contract, simplify, round corners, fill holes, then export to GeoPackage, Shapefile or GeoJSON\n\n\u2022 Installs itself: the model and its dependencies set up on first run, with no Python setup\n\n\u2022 Free tier with no card. Plans and limits: https://terra-lab.ai/pricing\n\n\u2022 Privacy: Semi-Auto on \"My computer\" sends nothing. \"Cloud AI\" sends a small image crop around each click to our servers. Automatic sends the imagery inside the zone you draw to our detection service\n\n\u2022 Windows, macOS and Linux, QGIS 3.22 to QGIS 4\n\n\u2022 12 languages: English, French, Portuguese, Spanish, German, Italian, Dutch, Polish, Indonesian, Japanese, Chinese (simplified and traditional)", "homepage": "https://terra-lab.ai/ai-segmentation", "repository": "https://github.com/TerraLabAI/QGIS_AI-Segmentation", "tracker": "https://github.com/TerraLabAI/QGIS_AI-Segmentation/issues", "author": "Yvann and Lilien from TerraLab", "tags": ["ai", "remote sensing", "machine learning", "deep learning", "satellite", "raster", "segmentation", "ia", "classification", "vectorization", "polygon", "digitizing", "artificial intelligence", "image segmentation", "feature extraction", "building detection", "object detection", "computer vision", "drone", "satellite imagery", "vector", "sentinel", "uav", "ortofoto", "remote-sensing", "polygons", "vegetation", "urban planning", "vetor", "landsat", "aerial", "orthophoto", "vector layer", "geotiff", "georreferenciamento", "neural network", "sentinel-2", "agriculture", "geopackage", "gpkg", "multispectral", "forestry", "instance segmentation", "satellite images", "buildings", "georeferencing", "segment anything"], "downloads": 68729, "latest_version": "3.3.1", "versions": [{"version": "3.3.1", "experimental": false, "qgis_min": "3.22.0", "qgis_max": "4.99.0", "downloads": 859, "uploaded_by": "yvannbarbot", "upload_datetime": "2026-10-07T04:53:25.675977"}]}