06-03, 13:30–15:00 (Europe/Stockholm), KO25 (WS4)
Modeling the spatial and temporal dynamics of natural and human-made processes often requires the development of complex models within Geographic Information Systems (GIS). As a result, this task has been limited to highly skilled GIS professionals. Our goal is to lower the entry barrier by developing an AI-powered modeling assistant called IntelliGeo, which facilitates the construction of geoprocessing workflows in the QGIS's Model Designer.
IntelliGeo is an open-source QGIS plugin that leverages the power of Large Language Models (LLMs) to make geospatial analysis more intuitive and efficient. It supports an iterative model-building process in which users define their requirements, refine them based on IntelliGeo’s outputs, and incorporate new insights gained through interactions with the plugin and testing the generated models.
To generate a model, IntelliGeo considers various factors, including user-defined requirements, available geoprocessing tools in QGIS, data loaded in the table of contents, and previously generated models. Currently, the model generation process is powered by general-purpose LLMs, which we instruct to produce geoprocessing models. However, user interactions with the LLMs will be used later to fine-tune an open-source LLM specifically for solving GIS-related problems.
At the QGIS User Conference 2024, we presented the initial steps of the plugin's development and received valuable feedback that helped steer our progress. Since then, we have made significant advancements, and in this workshop, we will showcase the evolution of IntelliGeo and demonstrate its current capabilities.
Workshop Agenda:
- Introduction to LLM Integration into GIS (10 minutes)
- Installation and connecting to LLM API Services (10 minutes)
- Step-by-Step Tutorial Using IntelliGeo (60 minutes)
- Wrap up and Q&A (10 minutes)
Gustavo García, a lecturer and researcher at the University of Twente's Faculty ITC, has a diverse academic background. Originally from Guatemala, he holds a bachelor's degree in software engineering and a master's in human resource management from his home country. In the Netherlands, he pursued further studies, obtaining a master's and PhD in geoinformatics. Prior to his current position, he accrued experience as a lecturer, software engineer, and researcher, notably focusing on geovisual analytics and collaborative analysis.
Mahdi is an Assistant Professor at the University of Twente, Faculty of Geo-Information Science and Earth Observation (ITC), Department of Geo-Information Processing (GIP) in Enschede, the Netherlands.
His research focuses on the intersection of artificial intelligence (AI) and geodata processing (GeoAI). His Ph.D. work involved using semantic web technologies and AI to automate geoprocessing workflows. Post-Ph.D., his research expanded to spatial data science and big data analytics, applying machine learning (ML) for spatiotemporal modeling in areas such as environmental monitoring, disaster management, and spatial epidemiology. He has also explored deep learning (DL) for extracting insights from large-scale spatial datasets and used natural language processing (NLP) for geospatial social media analysis.
His current research interests include:
• Large Language Models (LLMs) and Multimodal LLMs (MLLMs)
• MLOps and automated geoprocessing workflows
• Explainable AI (XAI)
• Spatial and spatiotemporal modeling using ML/DL
• Spatial data fusion