Deep Learning for Urban Land Cover
& Urban Land Valuation
Thematic Maps
Londrina-Maringá Metropolitan Region, Paraná, Brazil
Urban Land Cover Custom Classification
Deep Learning Models: CLIP RSICD ViTB/32 and DINOv2 ViTL/14 LoRA
Custom land cover classes: 10
Spatial Resolution: 109×109m
Year: 2024
Urban Land Value
Deep Learning Models: TabPFN v2 (1) and GAMLSS Spatio-Temporal (2)
(1) Predicted Unit Values: Median, Lower and Upper Bounds (CI=0.80) | Spatial Resolution: 109×109m | Year: 2024
(2) Predicted Unit Values: Median | Spatial Resolution: 50×50m | Time Period: 2000-2021
Bus Routes at Londrina
Year: 2025
Joint Analysis
Land Cover × Land Value Split View
Joint analysis of urban land cover classification (DINOv2 ViTL/14 LoRA) and median unit predicted urban land values (TabPFN v2) using a side-by-side slider visualization.
Left Layer: Urban Land Cover (10 classes) | Right Layer: Median Unit Land Value (R$/m²)
Spatial Resolution: 109×109m | Year: 2024
Visualize Urban Land Cover and Land Value Map
Urban Land Cover × Transport Structure
Joint analysis of urban land cover classification (DINOv2 ViTL/14 LoRA) and transport network infrastructure (OpenStreetMap functional hierarchy based on FHWA classification standards).
Visualize Urban Land Cover and Transport Structure Map
Urban Land Value × Transport Structure
Joint analysis of urban land value (TabPFN v2) and transport network infrastructure (OpenStreetMap functional hierarchy based on FHWA classification standards).
Visualize Urban Land Value and Transport Structure Map
Multimodal Urban Commute Reachability Analysis
Joint analysis of urban land value (TabPFN v2), land cover (DINOv2), transport network infrastructure (OpenStreetMap functional hierarchy based on FHWA classification standards) and commute reachability for 1 hour budget.
Visualize Multimodal Urban Commute Reachability Analysis Map
Institutional Notes
Research project conducted by Felinto Costa, PhD candidate in the UEM-UEL Joint Graduate Program in Architecture and Urbanism (PPU), affiliated with the Department of Statistics at the State University of Londrina (UEL), Brazil.
Research Team
- Author: Prof. MSc Eng. Felinto Costa – fjcosta@uel.br
- PhD Advisor: Prof. Dra. Mariana R. Urbano – mrurbano@uel.br
- Collaborator: Prof. Dr. Rodrigo R. Pescim – rrpescim@uel.br