AI-Driven Land Management: Revolutionizing Urban Planning
Published:
Project Overview
This project presents an innovative approach to urban planning by integrating state-of-the-art geospatial AI tools into land use analysis and development planning. It focuses on combining data-driven insights with scalable AI solutions to support sustainable urban growth and smarter city design.
Key Features
- Spectral Unmixing for identifying and classifying complex land cover types.
- Object Detection using Qwen2.5, applied to high-resolution satellite imagery.
- Visual Content Retrieval through LLAMA Index, enabling semantic image search and classification.
- Generative Urban Planning using Large Language Models (LLMs) to propose infrastructure layouts, zoning strategies, and public space designs.
Highlight
Most recently, I co-developed this geospatial AI prototype during the Sohjo Hackathon, where we integrated multiple modalities into a unified platform for urban planning experts and local authorities.
Technologies Used
- Python, PyTorch, Hugging Face Transformers
- LLAMA Index, LangChain
- Remote sensing datasets and spectral image processing
- Jupyter notebooks and Streamlit for prototyping
Outcome
The system was highly praised during the hackathon for its real-world applicability and innovative integration of AI with urban data. It demonstrates the potential of generative models in solving complex spatial problems and enhancing policy-making in urban contexts.