Trained deep learning models for street-level land surface temperature analysis using Google Street View imagery

This dataset contains the trained machine learning models used in the study “Differentiated street greenery cooling of road surfaces: Spatially varying nonlinear effects from street view and geographically weighted machine learning.” It includes (i) a global Random Forest model for land surface temperature prediction, (ii) a GeoShapley‑based Random Forest model for spatially explicit model interpretation, and (iii) a fine‑tuned Mask2Former semantic segmentation model for extracting street‑level vegetation and built‑environment features from Google Street View imagery. The paper was published in Sustainable Cities and Society: https://doi.org/10.1016/j.scs.2026.107501 All code and model pipelines are made open source on: https://github.com/toolambr/street_greenery