streetscapes is a Python package, CLI, and web-based explorer for large-scale analysis of street-level imagery.
It bundles functionality ranging from imagery retrieval to segmentation, feature extraction, and building-level aggregation.
The package is designed to be transparent, reproducible, and easy to extend for research use.
Requirements: Python 3.14
In a Python virtual environment, or conda environment do:
pip install streetscapes==1.0.0a0[sam3]To install for AMD GPUs (ROCm), do;
uv pip install streetscapes==1.0.0a0 --extra rocm,sam3Retrieving image metadata and downloading images is performed using the command line interface. For a simple quickstart, see the tutorial in the documentation. The documentation also contains more elaborate examples and workflows.
SAM3 requires the ultralytics package and "CLIP". However, CLIP is not on the Python packaging index. If for some reason ultralytics fails to install CLIP on its own, install it in your environment with:
pip install git+https://github.com/ultralytics/CLIP.gitIf you want to contribute to the development of streetscapes, have a look at the contribution guidelines.
streetscapes is licensed under CC-BY-SA-4.0.
This repository uses the data and work from the Global Streetscapes project.
[1] Hou Y, Quintana M, Khomiakov M, Yap W, Ouyang J, Ito K, Wang Z, Zhao T, Biljecki F (2024): Global Streetscapes — A comprehensive dataset of 10 million street-level images across 688 cities for urban science and analytics. ISPRS Journal of Photogrammetry and Remote Sensing 215: 216-238. doi:10.1016/j.isprsjprs.2024.06.023
The streetscapes package can be cited using the supplied citation information. For reproducibility, you can also cite a specific version by finding the corresponding DOI on Zenodo.