The Location Innovation Hub maintains a curated catalogue of cutting‑edge GeoAI tools and applications. These solutions show how artificial intelligence can enhance the use, analysis, and accessibility of geospatial data across sectors.
Our aim is to help organisations discover relevant tools and connect with the right providers.
Location Innovation Hub does not offer these solutions directly, but we act as a neutral matchmaker, making it easier to explore options, compare capabilities, and identify suitable partners for pilots or collaboration.
Interested in exploring GeoAI possibilities? Browse the solutions below and contact us if you want more information or a contact point to the solution provider.
Agriculture Analysis
AI solutions that enhance agricultural productivity using satellite imagery, drone data and field level sensors. These tools can support smarter decisions, yield prediction and sustainable farm management.
Organization:
Finnish Geospatial Research Institute, LUKE
Description of the AI solution:
AI and remote sensing were used to help silage (rehu) production management. AI predicts yield quantity and quality in silage swards.
Location data/tech:
- Remote sensing
- Multispectral imagery
Timeline:
Ended 31.10.2022 / -31.12.2022
Training data:
Drone remote sensing data sets from Jokioinen: RGB images, canopy height model (CHM), hyperspectral images (HSI). Produced by NLS and reference measurements by LUKE.
Training data license:
Not open publicly but could be for LIH customers in the future. Depends on Finnish Geospatial Research Institute and LUKE in the case of images and reference data. Related to LIH task 3.3.
Source codes:
Not open
More info:
https://www.maanmittauslaitos.fi/en/research/cybergrass
Publications:
Karila, K., Alves Oliveira, R., Ek, J., Kaivosoja, J., Koivumäki, N., Korhonen, P., ... & Honkavaara, E. (2022). Estimating Grass Sward Quality and Quantity Parameters Using Drone Remote Sensing with Deep Neural Networks. Remote Sensing, 14(11), 2692.
Organization:
Natural Resources Institute Finland
Description of the AI solution:
The EU's Green Deal sets clear targets for a more competitive and sustainable agriculture. This requires data-driven decision making for farmers, governments and policy makers.
The Horizon Europe ScaleAgData project aims to bridge the data gap of local-level observations by unlocking, integrating and upscaling in-situ farm sensor data.
Special focus is on innovations in sensor technologies, edge computing, data analytics and novel Earth observation products.
Luke leads the task for Data-based farming services. Work is based on a Digital Twin model of fields that integrates machinery, weather and remote sensing data with a biophysical simulation to support decision making. Methods are tested in cooperation with the Project's Research and Innovation Labs.
Location data/tech:
- Remote sensing
- Agricultural geodata
Timeline:
1.1.2023 – 31.12.2027
More info:
Organization:
Bayer Crop Science
Description of the AI solution:
Bayer uses AI algorithms to analyse satellite imagery to predict crop yields. Combining remote sensing data and machine learning enables accurate estimations that help farmers optimise operations.
Use Case: Agriculture and precision farming Benefits for Businesses: Supports planning, reduces waste, increases productivity and helps suppliers forecast demand. Headwords: Bayer Crop ScienceLocation data/tech:
- Satellite imagery
Timeline:
Ongoing
More info:
https://www.bayer.com/en/agriculture/ai-for-agriculture
https://www.bayer.com/media/en-us/bayer-pilots-unique-generative-ai-tool-for-agriculture/
Publications:
Bayer Crop Science, Agriculture Analysis, satellite imagery, Crop Yield Prediction Using Satellite Imagery and AI.
Environmental and Climate Analysis
GeoAI tools for understanding environmental change, analysing satellite‑based climate signals and improving resilience through data‑driven monitoring and forecasting.
Organization:
Finnish Geospatial Research Institute, University of Helsinki, VTT Technical Research Centre of Finland
Description of the AI solution:
Using the big data acquired with GNSS-Finland, the project develops new Deep Learning (DL) methods to compute future trends, detect signal anomalies, assess the continuity of location information, and forecast critical failures in positioning and timing information.
Location data/tech:
GNSS big data
Timeline:
1.9.2020 – 30.11.2023
Training data:
GNSS data from FinnPos stations (mostly GPS L1 and L2 signals). Openly available:
- https://finpos.nls.fi/gppstartpage_servlet/
- https://gnss-finland.nls.fi/#/map
- Also CODE (Center for Orbit Determination in Europe) data: ftp://gssc.esa.int/cddis/gnss/products/ionex/2021/
Training data license:
GeoGeo department maintains the core GNSS network. Hannu Koivula may know more.
Source codes:
Not public at the moment but may be opened as REASON is funded by the Research Council of Finland.
More info:
Publications:
- Kaasalainen, S., Mäkelä, M., Ruotsalainen, L., Malmivirta, T., Fordell, T., Hanhijärvi, K., ... & Nikolskiy, S. (2021). Reason – Resilience and security of geospatial data for critical infrastructures. In WiP Proceedings of the International Conference on Localization and GNSS (ICL-GNSS 2021). CEUR-WS.org.
- Morales Ferre, R., de la Fuente, A., & Lohan, E. S. (2019). Jammer classification in GNSS bands via machine learning algorithms. Sensors, 19(22), 4841.
- M. Saajasto, M. Mäkelä, F. S. Prol, M. Z. H. Bhuiyan and S. Kaasalainen (2023). Convolutional neural network based approach for estimating ionospheric delay from GNSS observables. ICL-GNSS 2023, Castellón, Spain. DOI: 10.1109/ICL-GNSS57829.2023.10148920.
Organization:
Clay Foundation
Description of the AI solution:
An open source AI model and interface for Earth.
Location data/tech:
Satellite imagery
Timeline:
2024–
Source codes:
https://github.com/Clay-foundation/model
More info:
Organization:
Google DeepMind
Description of the AI solution:
Foundation model used to produce the Google Satellite Embedding dataset, a global analysis-ready collection of learned geospatial embeddings.
Location data/tech:
- Satellite imagery
- Radar
- 3D point clouds
- Climate simulations
Timeline:
2025–
Source codes:
https://developers.google.com/earth-engine/datasets/catalog/GOOGLE_SATELLITE_EMBEDDING_V1_ANNUAL
More info:
https://deepmind.google/blog/alphaearth-foundations-helps-map-our-planet-in-unprecedented-detail/
Feature and Object Detection
AI‑powered methods for automatically detecting objects, features and changes in geospatial data, to boost efficiency in mapping, monitoring and situational awareness.
Organization:
Finnish Geospatial Research Institute
Description of the AI solution:
AI/deep learning interprets changes in aerial photos and laser scanning data, such as newly built or demolished buildings.
Location data/tech:
- Aerial photos
- Laser scanning data
Timeline:
Ended 2022, continued in AI4TDB
Training data:
- Building data digitised manually (Justus Poutanen), high quality; partly open, full dataset available on demand.
- Road data produced from Digiroad and NLS orthophotos (limited value).
- Hydrographic features (6 km × 6 km image) could be published – Juha and Tanja Heikkilä handle licensing.
Training data license:
- Building data partly open, no license.
- Road data not useful.
- Hydrography not open yet.
Source codes:
Not available
More info:
Publications:
- Koski, C., et al. (2022). Mapping small watercourses with deep learning. AGILE: GIScience Series, 3, 43.
- Koski, C., et al. (2021). Piloting ML for automatic mapping of streams. Abstracts of the ICA, 3, 1–2.
- Koski, C., et al. (2023). Mapping watercourses with deep learning. Remote Sensing, 15(11). https://doi.org/10.3390/rs15112776
Organization:
Finnish Geospatial Research Institute
Description of the AI solution:
AI/deep learning recognises hydrographical features using more versatile training data than ATMU.
Location data/tech:
- Hydrographical data
- Remote sensing
Timeline:
Started 2023
Training data:
Same as ATMU but expanded to streams, ditches and water areas.
Training data license:
Not yet available
Source codes:
Not yet available
Publications:
Not yet available
Organization:
Finnish Geospatial Research Institute, European Space Agency
Description of the AI solution:
AI automatically identifies objects such as navigation aids and vessels, improving situational awareness via sensor fusion.
Location data/tech:
- Sensor fusion
- Maritime imagery
Timeline:
Ended
Training data:
Visual imagery, sound recordings, RADAR, LiDAR, satellite navigation and vessel transponders.
Training data license:
Not open. ESA has IPR; NLS has usage rights. Contact Kristian Holmen.
Source codes:
ESA has the code – not public
More info:
Slides: NAVISP presentation
Publications:
- Thombre, S., et al. (2020). Sensors and AI techniques for autonomous ships: A review. IEEE Transactions on ITS, 23(1), 64–83.
Organization:
Finnish Geospatial Research Institute, European Space Agency
Description of the AI solution:
AI detects and identifies sea-ice features to support situational awareness for autonomous vessels.
Location data/tech:
Sea-ice remote sensing
Timeline:
Ended 30.4.2022, but continues
Training data:
Visual and infrared camera data with GNSS. Only partly labelled.
Training data license:
Most datasets cannot be open unless ESA decides otherwise.
Source codes:
To be clarified
More info:
Publications:
- Gorad, A., Hassan, S., & Särkkä, S. (2023). Vessel Bearing Estimation Using Visible and Thermal Imaging. SCIA 2023.
- Olkkonen, M.-K., et al. Maritime situational awareness in sea ice – Baška GNSS Conference (Extended Abstracts).
Organization:
Finnish Geospatial Research Institute
Description of the AI solution:
Master’s thesis by Emma Turkulainen: assessing DNNs to detect bark beetle infestations from UAS imagery at single-tree level.
Location data/tech:
UAS imagery
Timeline:
1.4.2019 – 31.12.2022
Training data:
High-resolution drone images (RGB, hyper- and multispectral) from forests (Paloheinä & Ruokolahti).
Training data license:
Not open (potentially open?)
Source codes:
Not available
More info:
https://www.maanmittauslaitos.fi/en/research/between-bark-and-wood-monituho
Publications:
- Kanerva, H., et al. (2022). Tree health decline from beetle damage using UAS RGB. Remote Sensing, 14(24), 6257.
- Turkulainen, E. (2023). DNN classification of spruce trees damaged by bark beetles.
Organization:
Natural Resources Institute Finland
Description of the AI solution:
AI service for automatic analysis of digital data – object detection, counting, classification of animals and more.
Location data/tech:
Aerial / satellite imagery
Timeline:
Ongoing
Training data:
Users upload their own photos
Training data license:
Crowdsourced data
More info:
Project: https://www.luke.fi/fi/projektit/aieye
Video: https://www.youtube.com/watch?v=4bbZ0vB4P88
Organization:
Finnish Geospatial Research Institute
Description of the AI solution:
Real-time wildfire smoke detection (segmentation) from drone imagery.
Location data/tech:
Drone imagery
Timeline:
Ongoing
Training data:
Manually labelled UAV smoke data + open datasets.
Training data license:
Partially available
Source codes:
Publications:
Pesonen, J., et al. (2024). Detecting Wildfires on UAVs with Real-time Segmentation. https://doi.org/10.48550/arXiv.2408.10843
Generic GeoAI
General purpose GeoAI libraries, models and frameworks that support machine learning workflows across a wide range of geospatial tasks and datasets.
Organization:
University of Tennessee
Description of the AI solution:
Python package for integrating artificial intelligence with geospatial data analysis and visualization.
Location data/tech:
- Satellite imagery
- Aerial photographs
- Vector data
Timeline:
2023–
Source codes:
https://github.com/opengeos/geoai
More info:
https://opengeoai.org/
Organization:
Wrocław Tech University
Description of the AI solution:
Python library for geospatial machine learning focusing on learning embeddings from vector geometries.
Location data/tech:
- Vector data (OSM, Overture Maps, GTFS)
Timeline:
2022–
Source codes:
https://github.com/kraina-ai/srai https://github.com/kraina-ai/srai
More info:
https://kraina-ai.github.io/srai/latest/ https://kraina-ai.github.io/srai/latest/
Organization:
TU Dortmund, Nvidia, Max Planck Institute, Stanford University
Description of the AI solution:
Graph Neural Network (GNN) library for PyTorch.
Location data/tech:
- Structured data (e.g. raster)
Timeline:
2019–
Source codes:
https://github.com/pyg-team/pytorch_geometric https://github.com/pyg-team/pytorch_geometric
More info:
https://pytorch-geometric.readthedocs.io https://pytorch-geometric.readthedocs.io
Organization:
OSGeo, TU Munich, Microsoft, University of Texas at San Antonio, University of Illinois Urbana-Champaign
Description of the AI solution:
TorchGeo provides datasets, samplers, transforms, and pre-trained models for geospatial data.
Location data/tech:
- Satellite imagery
- Aerial imagery
Timeline:
2021–
Source codes:
https://github.com/torchgeo/torchgeo https://github.com/torchgeo/torchgeo
More info:
https://www.osgeo.org/projects/torchgeo/ https://www.osgeo.org/projects/torchgeo/
Organization:
Azavea (USA)
Description of the AI solution:
Python library for deep learning on satellite and aerial imagery.
Location data/tech:
- Satellite imagery
- Aerial imagery
Timeline:
2018–
Source codes:
https://github.com/azavea/raster-vision https://github.com/azavea/raster-vision
More info:
https://rastervision.io/ https://rastervision.io/
Geoexploration
AI assisted approaches for exploring, visualising and interpreting geospatial datasets, from geological mapping to spatial pattern discovery.
Organization:
Finnish Geospatial Research Institute
Description of the AI solution:
- Knowledge on state‑of‑the‑art AI methods in geoexploration
- Determining suitable AI methodology for geoexplorative AI tools
Location data/tech:
- Geological datasets
- Remote sensing
Timeline:
Ended but may continue
Training data:
Does not exist
Training data license:
Does not exist
Source codes:
Does not exist
More info:
Overview: https://www.maanmittauslaitos.fi/en/research/tugeva-comprehending-unknown-geodata-through-geovisual-analytics-aided-artificial https://www.maanmittauslaitos.fi/en/research/tugeva-comprehending-unknown-geodata-through-geovisual-analytics-aided-artificial
Report: https://www.defmin.fi/files/5665/2500M-0134_TUGEVA_MATINE_Summary_Report_final.pdf https://www.defmin.fi/files/5665/2500M-0134_TUGEVA_MATINE_Summary_Report_final.pdf
Publications:
- Kettunen, P., Keskin, M. & Rönneberg, M. (2023). TUGEVA: Tuntemattoman paikkatietoaineiston ymmärtäminen tekoälyavusteisella geovisuaalisella analytiikalla. https://www.defmin.fi/files/5665/2500M-0134_TUGEVA_MATINE_Summary_Report_final.pdf
- Keskin, M. & Kettunen, P. (2023). Potential of eye‑tracking for interactive geovisual exploration aided by machine learning. International Journal of Cartography. https://doi.org/10.1080/23729333.2022.2150380
- Keskin, M., Rönneberg, M. & Kettunen, P. (2022). Cartographic adaptation through eye tracking and deep learning – GAIMS. EuroCarto 2022. https://doi.org/10.5194/ica-abs-5-107-2022
Organization:
Geological Survey of Finland
Description of the AI solution:
ArcSDM Spatial Data Modeler 5 for ArcGIS Pro.
Location data/tech:
Geophysical datasets
Timeline:
2016–2020, onwards
Training data:
https://github.com/gtkfi/demodata https://github.com/gtkfi/demodata
Training data license:
Public
Source codes:
Open
More info:
https://github.com/gtkfi/ArcSDM/wiki/Howto-start https://github.com/gtkfi/ArcSDM/wiki/Howto-start
Organization:
Geological Survey of Finland
Description of the AI solution:
Online modelling tool for mineral prospectivity.
Location data/tech:
Mineral prospectivity modelling
Timeline:
2014–2020
Training data:
- Basemap
- Geological and geophysical maps and datasets
- Mineral deposit data
- 3D modeller
Training data license:
Public
Source codes:
Not open
More info:
https://geodata.gtk.fi/mpm/ https://geodata.gtk.fi/mpm/
Geospatial Data Processing
Tools that automate and improve the processing, cleaning and transformation of geospatial data, to enable more accurate, efficient and scalable workflows.
Organization:
Finnish Geospatial Research Institute
Description of the AI solution:
AI produces land cover maps from satellite images. Automatic process based on laser scanning to generate labelled training data for a deep neural network.
Location data/tech:
- Satellite imagery
- Laser scanning
Timeline:
Ended 9/2022
Training data:
Masks (land cover maps) with classes: buildings, trees, low vegetation, low built-up areas, water. Includes satellite image patches.
Training data license:
Not open, but laser‑scanning‑based labels/masks may be shared in LIH after publication (subject to “Maanmittauslaitoksen Laserkeilausaineisto 5 p:n käyttöehdot”). Satellite imagery can be replaced with NLS ortho imagery.
Source codes:
Not open
More info:
https://www.maanmittauslaitos.fi/tutkimus/unite-metsien-ihmisten-ja-koneiden-vuorovaikutuksella-resilienssia-uusia-arvoverkkoja-ja https://www.maanmittauslaitos.fi/tutkimus/unite-metsien-ihmisten-ja-koneiden-vuorovaikutuksella-resilienssia-uusia-arvoverkkoja-ja
Publications:
- Karila, K. et al. (2023). Automatic labelling for semantic segmentation of VHR satellite images using airborne laser scanner data and OBIA. ISPRS Open Journal of Photogrammetry and Remote Sensing. https://doi.org/10.1016/j.ophoto.2023.100046
Organization:
Gispo
Description of the AI solution:
Tool for extracting cloud‑free mosaics from satellite data, developed for UN field operations. Pilot implemented with the UN Open GIS initiative for UNISFA in South Sudan.
Location data/tech:
Satellite imagery
Timeline:
2020
Training data:
Sentinel
Training data license:
Open
Source codes:
Open: https://github.com/GispoCoding/CFSI https://github.com/GispoCoding/CFSI
More info:
https://www.gispo.fi/en/blog/cloud-free-satellite-images-for-un-field-operations/ https://www.gispo.fi/en/blog/cloud-free-satellite-images-for-un-field-operations/
Organization:
Ubigu
Description of the AI solution:
Method for AI‑assisted identification and extraction of zoning regulations from PDF zoning documents. Enables full digitization and searchability of zoning regulations. Uses 3–4 different AI/ML techniques.
Location data/tech:
- Zoning documents
- Geospatial text extraction
Timeline:
To be specified
Organization:
Ubigu
Description of the AI solution:
AI-assisted method for optimizing land‑use implementation using multi‑criteria optimization. Selects optimal solutions under constraints (population, budget, location, quality, size). Generates scenario variations automatically. Applicable to land policy, property development, and infrastructure planning.
Location data/tech:
Geospatial planning data
Timeline:
To be specified
Organization:
Datafuchs
Description of the AI solution:
Natural‑language‑based geospatial analysis tool. Allows users to query spatial information without GIS expertise. Based on open data catalogue.
Location data/tech:
- Vector data (OSM, Overture Maps, GTFS)
Timeline:
2026–
More info:
https://www.datafuchs.com/geofuchs https://www.datafuchs.com/geofuchs
Real Estate Analysis
GeoAI applications for analyzing property data, detecting building‑related issues and supporting land valuation, planning and municipal decision‑making.
Organization:
Finnish Geospatial Research Institute
Description of the AI solution:
Automatic creation of land valuation using AI.
Location data/tech:
- Cadastral data
- Spatial modelling
Timeline:
To be specified
Training data:
- Cadastral index map
- Parcel price database
- Topographic database
- Sentinel satellite images
- NLS orthophotos
Training data license:
Not open
Source codes:
Not open
Organization:
Ubigu
Description of the AI solution:
AI‑based method for identifying unknown or unauthorized buildings in municipal and national registries, classifying them by various parameters. Uses national databases and/or satellite or aerial imagery to detect missing building information relevant for taxation or registry improvement.
Location data/tech:
- Aerial/satellite imagery
- Building registry data
Timeline:
To be specified
Organization:
Technical University Munich
Description of the AI solution:
Machine‑learning‑based pipeline for deriving building polygons and height information.
Location data/tech:
Satellite imagery
Timeline:
2025
Source codes:
https://github.com/zhu-xlab/GlobalBuildingAtlas https://github.com/zhu-xlab/GlobalBuildingAtlas
More info:
https://essd.copernicus.org/articles/17/6647/2025/ https://essd.copernicus.org/articles/17/6647/2025/
Other
Other AI resources ranging from computing to ecology and demographics.
Organization:
CSC – IT Center for Science Ltd.
Description of the AI solution:
Computing platform supporting numerous academic ML projects and commercial projects through LUMI. Provides specialist support for AI model execution and scaling. Also organizes the course “Practical machine learning for spatial data”.
Location data/tech:
- Spatial data processing
- HPC infrastructure
Timeline:
Ongoing
Source codes:
GeoML course code: https://github.com/csc-training/GeoML https://github.com/csc-training/GeoML
More info:
Examples: https://www.csc.fi/-/real-time-bird-monitoring-for-the-benefit-of-science https://www.csc.fi/-/real-time-bird-monitoring-for-the-benefit-of-science
https://www.csc.fi/-/towards-safer-navigation-and-fully-automated-vessels-with-ai-technologies https://www.csc.fi/-/towards-safer-navigation-and-fully-automated-vessels-with-ai-technologies
Organization:
University of Florida, University of Montpellier
Description of the AI solution:
Python package for deep learning on airborne RGB imagery, primarily used for tree detection.
Location data/tech:
Aerial imagery
Timeline:
2019–
Source codes:
https://github.com/weecology/DeepForest https://github.com/weecology/DeepForest
More info:
https://deepforest.readthedocs.io/ https://deepforest.readthedocs.io/
Organization:
University of Southampton
Description of the AI solution:
AI‑based global population data modelling using satellite imagery, statistics and land‑use datasets.
Location data/tech:
- Satellite imagery
- Population statistics
- Land‑use data
Timeline:
Ongoing, started 2013
Source codes:
https://github.com/wpgp https://github.com/wpgp
More info:
https://www.worldpop.org/ https://www.worldpop.org/