Article

Evaluate Spatial Equity in New York City’s Vibrant Street Programs with Local Lightweight Vision LLM (Working Paper)
Evaluate Spatial Equity in New York City’s Vibrant Street Programs with Local Lightweight Vision LLM (Working Paper)

This study evaluate multitemporal street improvements via a fine-tuned VLLM (Gemini 3).

Apr 1, 2025

Multi-Class Classification of Urban Regeneration Using a Siamese Network: An Analysis with Real-World Data from Portland, Oregon
Multi-Class Classification of Urban Regeneration Using a Siamese Network: An Analysis with Real-World Data from Portland, Oregon

This study proposes a deep learning framework using Siamese networks to classify urban regeneration into four categories based on high-resolution aerial imagery. Trained on 2,000 labeled parcels in Portland, Oregon, the model—especially with a U-Net backbone—achieved over 85% accuracy in detecting nuanced changes like redevelopment and demolition. The approach offers a scalable, cost-effective alternative to traditional data sources for monitoring urban regeneration.

Mar 2, 2025

LLMs Outperform Traditional NLP in Social Media Analysis for Urban Planning
LLMs Outperform Traditional NLP in Social Media Analysis for Urban Planning

This paper fine-tuned four SOTA LLM (GPT-4, Claude-3, Qwen2, and LLAMA3) for various social media analysis tasks, and results shows LLM outperform tradtional NLP methods in all areas.

Nov 25, 2023