Geospatial Artificial Intelligence
This course offers a comprehensive exploration of artificial intelligence in geography. Learn how AI technologies can identify and solve complex geographical issues
Practical information:
Want to register?
- Prerequisites: experience in Python programming and in handling geospatial data.
- Price: see tuition fees at Ghent University
You can enroll for this course via a credit contract at Ghent University.
Georganiseerd door:
Position of this course
In this course, you will delve into the fascinating integration of artificial intelligence (AI) within the field of geography. The course not only provides a comprehensive overview of the principles and concepts behind AI, but also enables you to gain practical experience through interactive applications with various AI technologies and algorithms. Moreover, you will learn how these advanced tools can be used to identify, analyze, and address complex geographical issues and challenges
Contents
- Overview and Introduction
- Overview of AI and its sub-domains
- Unsupervised & Supervised Learning | Traditional ML methods
- Example use cases of AI in the geospatial domain
- Ethical issues: bias and fairness
- Ontology (knowledge representation) and qualitative spatial reasoning
- Ontology
- Semantic web and linked Data
- Knowledge Graphs
- Qualitative Spatial Reasoning
- Hands-on: Python for (spatial) data cleaning, processing, and transformation
- Foundations of Artificial Neural Networks (ANNs)
- Basic neural network architecture
- Fitting a function with Stochastic Gradient Descent
- How a neural network approximates any function
- Multi-Layer Perceptron implementation
- Hands-on tutorials in Python & PyTorch
- Deep learning – Convolutional Neural Networks (CNNs)
- Computer vision problems in geography
- Feature learning versus feature engineering
- Convolutions, strides & padding, understanding convolution equations
- Hands-on training and implementation of a CNN on remote sensing data using fast.ai and PyTorch
- Deep learning – Recurrent Neural Networks (RNNs)
- RNNs (LSTM, GRU, Transformers) walk-through (basic ideas + hands-on tutorials)
- Applications of RNNs for geospatial applications: traffic prediction, next-location prediction
- Hands-on tutorials in Python
- Generative Adversarial Network (GAN) and Reinforcement Learning
- Introduction to NLP, LLM, “GenAI”, Hybrid-AI
- Project work with intermediate feedback
Final competences
- Understand the fundamentals of GeoAI and its main techniques and algorithms.
- Gain hands-on experience in various AI techniques and algorithms relevant to geospatial analysis using the Python programming language.
- Apply AI techniques to address or solve (simple) geographical problems and issues.
Teachers / speakers
Nico Van de Weghe
Prof. dr. Nico Van de Weghe is hoogleraar GIWetenschappen, een wetenschappelijke discipline op het snijvlak van informatica, sociale wetenschappen en natuurwetenschappen. GIScience houdt zich bezig met de studie van geospatiale informatie en hoe mensen de wereld begrijpen. Het behandelt ook hoe fenomenen van de wereld kunnen worden opgeslagen, gerepresenteerd en geanalyseerd. Het ligt aan de basis van GeoAI, dat als doel heeft machines te leren redeneren en ruimtelijk analyseren zoals een mens. Sinds zijn promotieonderzoek in 2004 heeft Van de Weghe een brede interesse in kennisgebaseerde AI, in het bijzonder tijdruimtelijke redenering en analyse.
Zoals op alle gebieden die met AI te maken hebben, worden datagestuurde benaderingen steeds dominanter in GeoAI. Daarom richt Van de Weghe zich momenteel op onderzoek in zowel kennisgebaseerde als datagestuurde GeoAI. Hij onderzoekt de mogelijkheden om beide te combineren tot hybride GeoAI. Sinds zijn vroege onderzoeksperiode ligt Van de Weghe's nadruk op tijdruimtelijke toepassingen in het algemeen en de analyse van bewegende objecten in het bijzonder. Hij heeft zijn onderzoek toegepast in verschillende domeinen, waaronder diergedrag, criminologie, massa-analyse, gezondheidszorg, mobiliteitsonderzoek, navigatie en sportanalyse.
Expertise: GeoAI, Generatieve AI, GIWetenschappen, Geografische informatiesystemen (GIS), kennisgebaseerd AI, datagebaseerd AI, hybride AI
Haosheng Huang
Haosheng Huang is a Professor in GIScience and Cartography at Ghent University. He obtained a BSs/MSc degree (with distinction) in Computer Science and a PhD degree (with distinction) in Geoinformation/Cartography. Previously, he was a senior lecturer and research group leader at the Department of Geography at the University of Zurich, Switzerland (2016-2020); a researcher and lecturer (Univ.Ass.Dr.techn.) at Research Group Cartography, TU Wien, Austria (2013-2016).
Haosheng has a broad research interest in the modelling, analysis, and visualization of geographical data, as well as their interdisciplinary applications. His research expertise includes: Geospatial AI (GeoAI), Spatial Data Science, Urban informatics, Location based services and mobile HCI (e.g., augmented reality AR), and Spatial cognition (e.g., with eye-tracking and EEG). The application domains range from navigation systems (for pedestrians, cyclists, and car drivers), sustainable mobility, and smart cities to social equity and health sciences.
Expertise: Geospatial AI (GeoAI), Spatial Data Science, Virtual Reality (VR), Mobile HCI, Smart Cities and Mobility
Tim Van de Voorde
In my research, remote sensing, spatial analysis and GIS techniques are used to help cities meet the challenges they face posed by rapid urbanisation, unplanned growth and global environmental change.
My work strives to combine aspects of both fundamental and applied research while addressing societal challenges, supporting policy development and including valorisation. I have developed methods and applications to map urban green infrastructure, monitor and model land-use change, urban growth and the resulting environmental impacts. This includes, for example, forecasting the spatial impact of alternative planning scenarios for strategic environmental impact assessment, improving hydrological models that predict runoff in urban watersheds to assess flood risk, and developing indicators of urban vegetation abundance and accessibility of urban green spaces.
My expertise includes:
- Urban green infrastructure and ecosystem services
- Urban remote sensing
- Sub-pixel classification of medium resolution satellite data
- Land-use change models
- Geographic information systems
- Python, ArcGIS and open-source GIS solutions
- Machine-learning approaches in remote-sensing based image classification
Related courses
HPC Café
Seminarie - Online, Karlsruhe - KIT, HammerHAI, Training geselecteerd in de Belgische AI Factory Antenna (AIFA)
