GeoAI

The Invisible City: Not All Streets Are Seen Equally
Investigates spatial bias and systematic omission in urban street view datasets and digital representations.
Mapping the Semantics of the Street: A VLM-Driven Geospatial Analysis in Fatih, Istanbul
Deploys Vision-Language Models on street-level imagery across Fatih, Istanbul, extracting multi-attribute urban semantics and mapping vibrancy.
DOI
A Graph-Based Community Detection Approach for Identifying the Semantic Neighbourhoods within London's Airbnb Properties
Awarded the Best Short Paper Award at the 29th AGILE Conference. Delineates semantic urban neighbourhoods in London using graph community detection on property listings.
DOI
Emergent Spatio-Semantic Structure in Large Language Model Embedding Spaces
Demonstrates emergent geospatial and spatial-semantic structure within LLM embedding spaces without explicit coordinate training.
DOI
Towards GeoAI Foundation Models: A Multimodal Learning Framework with Geospatial Intelligence
Conceptual and architectural framework for building multimodal GeoAI foundation models with geospatial intelligence.
Cite
Digitalising Probation with Geospatial AI: Street-Level Human Perception in İstanbul’s Fatih District
Applies Geospatial AI and street-level perception analytics to modernize probation monitoring and spatial risk assessment in Istanbul.
Street-Level Geolocalization Using Multimodal Large Language Models and Retrieval-Augmented Generation
Proposes a multimodal framework integrating Vision-Language Models with RAG for street-level image geolocalization and interpretable spatial reasoning.
DOI
Yapay Zekanın Sokak Seviyesi Görüntü Analizinde Kullanımı
Sokak seviyesi görüntülerin yapay zeka, bilgisayarla görme ve VLM teknolojileriyle analiz edilmesine yönelik yöntem ve uygulamalar.
ATTransUNet: Semantic Segmentation Model for Building Segmentation from Aerial Image and Laser Data
Proposes ATTransUNet, combining attention mechanisms and transformers for high-precision building segmentation from aerial and LiDAR data.
DOI