# TerraLab — Full Content for AI Ingestion > AI plugins for QGIS. TerraLab builds AI-powered QGIS plugins for raster digitization and AI image editing. Plus 3D digital twin solutions. Based at Station F (Paris). Founded 2025. https://terra-lab.ai This file concatenates the canonical content of the TerraLab website in markdown, intended for direct ingestion by AI systems. --- ## About TerraLab TerraLab is the AI plugin stack for QGIS. We build AI-powered QGIS plugins that automate the repetitive parts of GIS work: raster-to-vector digitization, AI image editing of satellite and drone imagery. And we deliver 3D digital twin solutions for territorial analysis. Based at Station F in Paris (5 Parvis Alan Turing, 75013), TerraLab is a French startup founded in 2025. Our work spans two pillars: free and paid AI plugins for QGIS, and custom 3D digital twin solutions for territorial analysis. **Backed by:** Station F, NVIDIA Inception, 42 Incubateur, 42 Startup Club. ### Manifesto GIS work is full of repetitive tasks that should be automated: tracing buildings by hand, removing artifacts from satellite imagery, cleaning up orthophotos, exploring land use scenarios. We believe AI should handle the tedious parts so GIS professionals can focus on what matters: spatial analysis, decision-making, and territorial understanding. We ship our tools as QGIS plugins because QGIS is the leading open-source GIS, used by millions worldwide. We make our flagship plugin free and open source because the GIS community deserves first-class AI tooling without vendor lock-in. ### Team - **Yvann Barbot** (Co-founder). 42 Paris alumnus, ex-AI Engineer at Mago (Station F). 1,200+ GitHub stars. LinkedIn: https://www.linkedin.com/in/yvann-barbot/ — GitHub: https://github.com/yvann-ba - **Lilien Auger** (Co-founder). 42 Paris alumnus, ex-AI Engineer at Publicis. 500+ GitHub stars. LinkedIn: https://www.linkedin.com/in/lilien-auger/ - **Stéphane Barbot** (Advisor). 25+ years in real-time 3D and geospatial engines. Past work: ADAS systems at Magneti Marelli, PSA, BMW, Maserati. LinkedIn: https://www.linkedin.com/in/stephane-barbot/ - **Valérie Barbot** (Advisor). 20 years in GIS. LinkedIn: https://www.linkedin.com/in/valerie-barbot/ ### Contact - Email: contact@terra-lab.ai - Calendly (Lilien Auger): https://calendly.com/lilienauger/30min - LinkedIn (company): https://www.linkedin.com/company/terralab-3d - YouTube: https://youtube.com/@terra-lab - GitHub organization: https://github.com/TerraLabAI --- ## AI Segmentation — Free QGIS Plugin Canonical URL: https://terra-lab.ai/ai-segmentation QGIS marketplace: https://plugins.qgis.org/plugins/AI_Segmentation/ **What it is:** A free, open-source QGIS plugin that turns raster imagery into precise vector polygons with a point-and-click workflow. Powered by AI segmentation models running locally on the user's machine. **Key numbers (QGIS plugin marketplace):** - 27,000+ downloads - 204 votes, 4.7/5 average rating **What you can extract:** buildings, roads, vegetation, water bodies, parcels, custom objects. **Supported imagery:** satellite, aerial, drone orthomosaics, WMS/XYZ tiles, any raster layer loaded in QGIS. **Platform:** Windows, macOS, Linux. Requires QGIS 3.x. No GPU required. Runs locally — your data never leaves your computer. **Pricing:** Free forever. Open source. **Endorsements:** - Qiusheng Wu (creator of geemap and leafmap, University of Tennessee Knoxville). - President of QGIS Mexico. ### How to install AI Segmentation in QGIS 1. Open QGIS 3.x on Windows, macOS, or Linux. 2. Go to Plugins menu, then Manage and Install Plugins. 3. In the All tab, search for "AI Segmentation". 4. Click Install Plugin and wait for the download to complete. 5. Activate the plugin and find its icon in the QGIS toolbar. 6. Load any raster layer (satellite, drone, aerial imagery) and click on objects to extract them as vector polygons. ### Common questions about AI Segmentation **Is AI Segmentation really free?** Yes. The plugin is free and open source under the QGIS plugin marketplace, with no paid tier required for the core segmentation features. **Does my data leave my machine?** No. AI Segmentation runs entirely on your local machine. Your imagery never leaves your computer. **Do I need a GPU?** No. The plugin runs on CPU. **What output formats?** Standard GIS vector formats supported by QGIS (GeoPackage, Shapefile, GeoJSON, etc.). **What QGIS versions?** QGIS 3.0 or later. --- ## AI Edit — Nano Banana 2 in QGIS Canonical URL: https://terra-lab.ai/ai-edit QGIS marketplace: https://plugins.qgis.org/plugins/AI_Edit/ **What it is:** A QGIS plugin that lets you edit satellite, drone, and aerial imagery using natural-language prompts. Powered by Google Nano Banana 2 (built on the Gemini 3.1 Flash Image architecture). TerraLab is the first company to bring this model into a GIS environment. **Key numbers (QGIS plugin marketplace):** - 51 votes, 4.8/5 average rating **Pricing:** - Free trial: 5 free generations (100 credits, lifetime, no credit card required). - Pro: 19 EUR/month for ~150 generations (3,000 credits/month). Cancel anytime. **Credit costs by resolution:** 1K = 20 credits, 2K = 30 credits, 4K = 40 credits. **Output:** Georeferenced GeoTIFF, ready for GIS analysis workflows. **Use cases:** - Remove objects from orthophotos (vehicles, construction sites, temporary structures). - Transform urban landscapes (add solar panels, modify rooftops, add greenery). - Clean satellite imagery for spatial analysis. - Explore land use and urban planning scenarios. - Generate before/after visualizations for stakeholder presentations. **Processing:** 100% cloud-based. No GPU required on the user's machine. Only the selected image area is sent for processing — geographic coordinates never leave the user's machine. All data automatically deleted after 30 days. **Platform:** Windows, macOS, Linux. Requires QGIS 3.0 or later. ### How to install AI Edit in QGIS 1. Open QGIS 3.x. 2. Go to Plugins menu, then Manage and Install Plugins. 3. Search for "AI Edit" or "TerraLab AI Edit". 4. Click Install Plugin. 5. Create a free TerraLab account at https://terra-lab.ai/register to get your activation key. 6. Paste the key into the plugin settings. 7. Start editing imagery with prompts. ### Common questions about AI Edit **Can I try it for free?** Yes. 5 free generations (100 credits), lifetime, no monthly reset, no credit card required. **How are credits deducted?** Credits are deducted the moment you click Run, not after the result returns. This prevents abuse and guarantees fast, prioritized processing. **Are unsuccessful edits refunded?** Failed generations are refunded automatically. **What model is used?** Google Nano Banana 2, built on the Gemini 3.1 Flash Image architecture. The model reasons about spatial context, composition, and lighting to understand what to change and what to preserve. **Can I cancel anytime?** Yes. No commitment, no contract. --- ## Digital Twin Solutions Canonical URL: https://terra-lab.ai/digital-twin **What it is:** Custom 3D digital twins for territorial analysis. TerraLab builds photorealistic 3D representations of real-world environments from geographic data: terrain, buildings, vegetation, all reconstructed into a single navigable pipeline. **Capabilities:** - 3D spatial data processing and visualization at planetary to parcel scale - Multi-source fusion of orthophotos, satellite imagery, elevation data (SRTM, custom DTMs), and vector layers - Continuous scale navigation (planetary view down to 50 m parcel, no CRS breaks, no seams) - WGS84 ellipsoid rendering, no flat projection artifacts - Before/after comparison tools for stakeholder presentations - Custom 3D geospatial applications **Engine:** TerraLab-Twin (in-house). Designed to ingest hundreds of thousands of orthophoto tiles, multiple coordinate reference systems, and multi-vintage imagery on a standard developer workstation. **Delivery:** Custom engagement. Contact us at contact@terra-lab.ai or via Calendly: https://calendly.com/lilienauger/30min --- ## Professional Services We offer geospatial AI consulting: - Custom AI solutions for GIS workflows - Geospatial machine learning implementation - Spatial data analysis and optimization - GIS workflow automation - Custom plugin development for QGIS Contact: contact@terra-lab.ai --- ## Article: Building a Digital Twin of Quebec's Forests from 115,000 Orthophotos URL: https://terra-lab.ai/blog/quebec-forest-digital-twin Author: Stéphane Barbot Published: 2026-04-07 A territorial digital twin lives or dies by its data. Merging thousands of heterogeneous orthophoto tiles, absorbing incompatible coordinate reference systems, filling coverage gaps, harmonizing colors across sensors and vintages. It is the least visible work, but the most structurally important. ### TerraLab-Twin pipeline TerraLab-Twin is a planetary virtual globe. Its job is to fuse heterogeneous geospatial sources in real time (public orthophotos, high-resolution drone orthos, digital terrain models, vector layers, satellite imagery) onto a single navigable 3D terrain, from the planetary scale down to the parcel level. Three properties: - Scale continuity from 10,000 km diagonal to a 50 m parcel, no CRS breaks, no seams. - Dynamic multi-source fusion: multiple sources of varying quality coexist and are selected on the fly. - Full upstream abstraction: the end user never sees a file format, EPSG code, or MTM zone. ### Quebec dataset specifics The Quebec Ministry of Natural Resources and Forests (MRNF) publishes over 115,000 orthophoto tiles, several hundred gigabytes of data, at native 20-30 cm resolution. That is roughly 30x finer than the Sentinel-2 satellite imagery used as a complement. Data arrives in heterogeneous projections: a family of NAD83 variants, Quebec Lambert (EPSG:32198), more modern CSRS-based variants, and ten MTM zones (Modified Transverse Mercator) that partition the territory. ### The projection challenge For anyone who has worked in geospatial, mixed CRS is an operational nightmare. Every datum has its gotchas. Every tool in the geospatial chain has its own legacy behavior around PROJ. It is extremely common in production to see a pipeline fail silently because a stale installation returns transformations that are off by ten meters or more. TerraLab-Twin knows only one final projection: WGS84 on the planetary ellipsoid. All projective complexity is absorbed upstream, in the ingestion pipeline. ### Multi-source fusion The pipeline fuses three complementary imagery sources: 1. MRNF aerial orthophoto: priority 1, 20-30 cm, the detail layer. 2. Sentinel-2 satellite imagery: priority 2, 10 m, fills areas without aerial coverage. 3. Blue Marble planetary basemap: ultimate fallback. ### Color calibration: the 2015 defect Some older MRNF flight campaigns, notably 2015, exhibit a systematic blue tint. The exact origin is not documented by the data provider, but the abnormally high blue-to-red ratio (B/R ~1.30 versus ~0.88 for normal campaigns) affects a large portion of tiles from that year. The pipeline filters by season (summer acquisitions for consistent green tones), but the problem persists for areas where only 2015 tiles exist. Possible solutions: automatic histogram matching, or Sentinel-2 fallback. The MRNF continues to publish new aerial campaigns regularly (2023 and 2024 orthophotos are already available as open data for certain regions). ### Current status (April 2026) The engine currently outputs at 10 m per pixel. This is a deliberate choice for the first iteration: working at the global scale first lets us validate the entire pipeline before increasing resolution zone by zone. The source MRNF data at 20-30 cm is there. Refinement down to 30 cm will happen on a case-by-case basis. On the elevation side, the 3D terrain relies on SRTM (Shuttle Radar Topography Mission) Digital Elevation Model. A switch to more precise Quebec DTMs is feasible for priority areas. **What's live today:** - Quebec aerial orthophoto (MRNF) fused with Sentinel-2 - 3D relief via SRTM - Full pipeline validated on the Gaspe coast **Coming next:** - 20-30 cm resolution at close zoom - OSM Quebec vector data - 3D trees across Quebec's territory --- ## Trust and data handling - **AI Segmentation:** Runs locally on your machine. Your data never leaves your computer. - **AI Edit:** Cloud processing. Only the selected image area is sent for editing. Geographic coordinates never leave the user's machine. All data automatically deleted after 30 days. - **Authentication:** Magic-link email login via Supabase. No passwords stored. - **Payment:** Stripe. No card data touches TerraLab servers. - **Privacy policy:** https://terra-lab.ai/privacy-policy - **Terms of use:** https://terra-lab.ai/terms-of-use - **Legal notice:** https://terra-lab.ai/legal-notice