SAGAI v2.0: streetscape analysis with AI in a single Google Colab notebook

Streetscape analysis with AI: score and map every street of a city, from a bounding box to street-level indicators, by asking questions in plain language, with no training data.

What is SAGAI?

SAGAI is an open-source, zero-shot workflow designed to score and map street-level urban environments using vision-language models. With no need for labeled datasets or deep learning expertise, SAGAI lets users analyze streetscapes through natural language prompts using nothing more than a geographic bounding box.

SAGAI v2.0 runs as a single Google Colab notebook that handles the entire pipeline, from street sampling to image retrieval, VLM-based scoring and thematic mapping. It integrates UVLM (Unified Vision-Language Model Loader) directly as its inference engine, giving access to 24 VLM checkpoints (7 LLaVA-NeXT, 4 Qwen2.5-VL, 4 Qwen3-VL, 6 InternVL3.5 and 3 Gemma 4 models), from 1B to 110B parameters, all through a unified interface.

You can:

  • Classify environments (e.g., urban vs. rural)
  • Detect features (e.g., storefronts, sidewalks)
  • Estimate physical attributes (e.g., sidewalk width)
  • Benchmark multiple VLMs on the same task with identical prompts
  • Design any custom visual analysis task by adapting the prompt

All you need is a location: no pretraining, no fine-tuning and no annotation.

SAGAI in under a minute


How It Works

SAGAI v2.0 runs as a single Google Colab notebook. The pipeline executes six stages sequentially:

The six blocks of the SAGAI notebook, from street sampling to maps

1. Street Sampling Using OpenStreetMap, SAGAI automatically extracts the pedestrian street network within a user-defined bounding box and generates evenly spaced sampling points along it.

2. Image Retrieval Google Street View images are downloaded at each sampling point in multiple compass directions using the Street View Static API.

3–5. VLM Scoring (powered by UVLM) Each image is scored using a vision-language model selected by the user. UVLM abstracts away all architecture-specific inference details and provides a single unified function that works identically across model families. Users choose a model, set a quantization level, and define their analysis tasks through a prompt builder.

6. Aggregation and Mapping Scores are aggregated at point and street segment levels using GeoPandas and visualized as thematic maps ready for publication or further spatial analysis.

SAGAI architecture: geospatial data, VLM inference through UVLM, and visualization

Supported Models

Through UVLM, SAGAI v2.0 supports 24 VLM checkpoints from five model families: 7 LLaVA-NeXT, 4 Qwen2.5-VL, 4 Qwen3-VL, 6 InternVL3.5 and 3 Gemma 4 models, from 1B to 110B parameters.

With 4-bit quantization, models up to 14B run on Colab’s free T4 GPU, and models up to 34B on an A100. Larger models need a multi-GPU environment.

Scoring Tasks

SAGAI ships with predefined scoring tasks and supports four response types through UVLM’s prompt builder:

NumericAn integer or a decimal, e.g. sidewalk width in meters
CategoryA label among options, e.g. vegetation type
BooleanYes or no, e.g. storefront presence
TextA free-form description

All prompts are fully customizable. Any visual analysis task can be defined through prompt modification alone: building typology, cycling infrastructure, pedestrian comfort, accessibility, street furniture, and more.

Key Capabilities

Single notebookThe full pipeline from bounding box to thematic maps in one place
Multi-model inferenceBenchmark and compare VLMs on identical tasks using identical prompts
Consensus validationMajority voting across 2–5 repeated inferences per image for improved reliability
Chain-of-thought reasoningAdjustable token budget (up to 1,500 tokens) for complex visual reasoning tasks
Truncation detectionAutomatic flagging of responses that hit the generation limit, with per-task diagnostics
Resume-safe batch processingCheckpoint saving every 3 images, automatic resume on interruption
Quantization support4-bit, 8-bit, and FP16 precision via BitsAndBytes
Publication-ready outputsCSV scores, GeoPackages, and thematic maps at point and street segment levels

Why SAGAI?

100% open-source (Apache 2.0)
Zero-shotNo training data, no fine-tuning, no annotation
Multi-model24 VLMs through UVLM, not locked to a single architecture
Runs entirely in Google ColabNo local installation, no GPU of your own needed
Prompt-drivenAdapt to any urban analysis task by changing the prompt
Publication-readyGenerates maps, scores, and spatial datasets out of the box

Get Started

Open the notebook in Google Colab and follow the steps. You need a Google account, and a Google Maps API key with the Street View Static API enabled to download the images.

Citations & Publications

If you use SAGAI in your work, please cite:

Perez, J. & Fusco, G. (2026) ‘From Street View Imagery to Street Quality Indicators: Vision Language Inference for the Suburban 15-minute City’, arXiv preprint arXiv:2608.20026. Available at: https://arxiv.org/abs/2608.20026

Perez, J. & Fusco, G. (2025). Streetscape Analysis with Generative AI (SAGAI): Vision-Language Assessment and Mapping of Urban Scenes. Geomatica, Volume 77, Issue 2, 100063.
DOI: 10.1016/j.geomat.2025.100063

Read the publication

Assessing urban scenes for the 15-minute city through SAGAI (Streetscape Analysis with Generative AI) – Presentation – 24th European Colloquium on Theoretical and Quantitative Geography

PDF of the presentation


Funding & Partners

SAGAI is developed within emc2 (Evolutive Meshed Compact City), a research project of the Driving Urban Transitions (DUT) Partnership, co-funded by the European Union, with the support of CNRS.

emc2, Evolutive Meshed Compact CityDUT, Driving Urban TransitionsCNRSCo-funded by the European Union

© 2025–2026 Joan Perez (Urban Geo Analytics) and Giovanni Fusco (CNRS). SAGAI is open source, released under the Apache 2.0 license.

Banner: aerial footage “Marseille par drone” by SEVENBELL PRODUCTION (CC BY 3.0, via Wikimedia Commons), darkened, with the street network © OpenStreetMap contributors (ODbL).