Dual Friction Mapping of Urban Pedestrian Spaces — Data

<p dir="ltr">Supplementary data archive for: Eminoğlu, Y., Çubukçu, K.M. (2026). Dual friction mapping of urban pedestrian spaces: Integrating street-view visual analytics and Geographically Neural Network Weighted Regression in Izmir, Türkiye. Sustainable Cities and Society, 107913. doi.org/10.1016/j.scs.2026.107913</p><p><br></p><p dir="ltr">Code and analysis scripts are maintained separately in the public companion repository: gitlab.com/yusufeminoglu/pfi_paper (mirrored at github.com/YusufEminoglu/pfi_paper).</p><p><br></p><p dir="ltr">This archive contains: (1) six GeoPackage layers (EPSG:4326): pfi_merged (5,663 street-view points with the Pedestrian Friction Index, its components and predictors), gnnwr_results (5,663 points with GNNWR local coefficients, predictions and residuals), hex_all (8,191 H3 hexagons), hex_clustered (397 eligible hexagons with GMM morphological clusters), hex_sampled (126 sampled hexagons) and sam3_hex_summary (126 sampled hexagons with hexagon-level visual metrics); (2) processed_tables_csv.zip, the CSV tables of the analysis pipeline (merged PFI dataset, GNNWR results, SegFormer-B5 visual metrics and SAM 3 validation scores, Street View node list, thermal/park-access metrics and street-network metrics); (3) gsv_validation_sample.zip, a small illustrative sample (10 nodes x 4 directions = 40 images) of raw Google Street View crops and their SegFormer-B5/ADE20K segmentation overlays used for the SAM 3 vs. SegFormer-B5 validation exercise reported in the article.</p><p><br></p><p dir="ltr">The full Google Street View image corpus (5,663 points x 4 directions) is not redistributed here due to third-party (Google) licensing restrictions on bulk re-publication of Street View imagery; only the small illustrative validation subset above is included.

Node-level metadata (node ID, coordinates, hexagon ID, district, morphological cluster) for the full corpus is included in processed_tables/sampling/gsv_nodes.csv. Additional research materials may be requested from the corresponding author.</p><p><br></p><p dir="ltr">License: the CC BY 4.0 license of this dataset applies only to material created by the authors (the GeoPackage layers and CSV tables); it does not apply to third-party Google Street View imagery in gsv_validation_sample.zip.