Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Streetscape Analysis with Generative AI (SAGAI): Vision-Language Assessment and Mapping of Urban Scenes

Joan Perez, Giovanni Fusco

TL;DR

The paper tackles scalable streetscape assessment by leveraging vision-language reasoning with open geospatial data. It introduces SAGAI, a four-module zero-shot workflow that links OpenStreetMap, Google Street View, and LLaVA to generate structured urban indicators at point and street scales. Empirical evaluation in Nice and Vienna shows high accuracy for urban/rural classification, moderate storefront detection, and informative sidewalk-width estimates, highlighting both promise and current limitations of zero-shot VLMs in urban contexts. The open-source, Colab-ready toolkit enables researchers and planners to tailor prompts for walkability, safety, or urban design analyses, supporting reproducible, scalable analysis of streetscapes.

Abstract

Streetscapes are an essential component of urban space. Their assessment is presently either limited to morphometric properties of their mass skeleton or requires labor-intensive qualitative evaluations of visually perceived qualities. This paper introduces SAGAI: Streetscape Analysis with Generative Artificial Intelligence, a modular workflow for scoring street-level urban scenes using open-access data and vision-language models. SAGAI integrates OpenStreetMap geometries, Google Street View imagery, and a lightweight version of the LLaVA model to generate structured spatial indicators from images via customizable natural language prompts. The pipeline includes an automated mapping module that aggregates visual scores at both the point and street levels, enabling direct cartographic interpretation. It operates without task-specific training or proprietary software dependencies, supporting scalable and interpretable analysis of urban environments. Two exploratory case studies in Nice and Vienna illustrate SAGAI's capacity to produce geospatial outputs from vision-language inference. The initial results show strong performance for binary urban-rural scene classification, moderate precision in commercial feature detection, and lower estimates, but still informative, of sidewalk width. Fully deployable by any user, SAGAI can be easily adapted to a wide range of urban research themes, such as walkability, safety, or urban design, through prompt modification alone.

Streetscape Analysis with Generative AI (SAGAI): Vision-Language Assessment and Mapping of Urban Scenes

TL;DR

The paper tackles scalable streetscape assessment by leveraging vision-language reasoning with open geospatial data. It introduces SAGAI, a four-module zero-shot workflow that links OpenStreetMap, Google Street View, and LLaVA to generate structured urban indicators at point and street scales. Empirical evaluation in Nice and Vienna shows high accuracy for urban/rural classification, moderate storefront detection, and informative sidewalk-width estimates, highlighting both promise and current limitations of zero-shot VLMs in urban contexts. The open-source, Colab-ready toolkit enables researchers and planners to tailor prompts for walkability, safety, or urban design analyses, supporting reproducible, scalable analysis of streetscapes.

Abstract

Streetscapes are an essential component of urban space. Their assessment is presently either limited to morphometric properties of their mass skeleton or requires labor-intensive qualitative evaluations of visually perceived qualities. This paper introduces SAGAI: Streetscape Analysis with Generative Artificial Intelligence, a modular workflow for scoring street-level urban scenes using open-access data and vision-language models. SAGAI integrates OpenStreetMap geometries, Google Street View imagery, and a lightweight version of the LLaVA model to generate structured spatial indicators from images via customizable natural language prompts. The pipeline includes an automated mapping module that aggregates visual scores at both the point and street levels, enabling direct cartographic interpretation. It operates without task-specific training or proprietary software dependencies, supporting scalable and interpretable analysis of urban environments. Two exploratory case studies in Nice and Vienna illustrate SAGAI's capacity to produce geospatial outputs from vision-language inference. The initial results show strong performance for binary urban-rural scene classification, moderate precision in commercial feature detection, and lower estimates, but still informative, of sidewalk width. Fully deployable by any user, SAGAI can be easily adapted to a wide range of urban research themes, such as walkability, safety, or urban design, through prompt modification alone.

Paper Structure

This paper contains 12 sections, 6 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology: The SAGAI Workflow
  4. Module 1: OSM Point Generator
  5. Module 2: Street View Batch Downloader
  6. Module 3: Scene Assessment with LLaVA
  7. Module 4: Geospatial Scoring Aggregation and Mapping
  8. Open-Source Availability and Repository Access
  9. Empirical Evaluation: Two Case Studies in the Peripheral Areas of Nice and Vienna
  10. Experimental Results
  11. Discussion and Perspectives
  12. Conclusion

Figures (6)

  • Figure 1: The Four Modules of the SAGAI Workflow
  • Figure 2: Street View Image Availability by Sample Point in Nice (left) and Vienna (right)
  • Figure 3: Examples of Misclassified Images with Highlighted Interpretation Gaps between Human and AI Scoring
  • Figure 4: Averaged Urban Character Scores from Module 4 – Task 1 (Categorization) for Nice (left) and Vienna (right), at Point and Street Segment Levels
  • Figure 5: Total predicted scores for storefront presence at the point level (left subpanels) and aggregated by street segment (right subpanels) from Module 4 – Task 2 (Counting) shown for Nice (top) and Vienna (bottom)
  • ...and 1 more figures