pathways-bench 0.0.1

A Benchmark for Extracting Routable Pedestrian Path Network Graphs

PathwayBench: A Benchmark for Extracting Routable Pedestrian Path Network Graphs

This package contains the PathwayBench dataset and benchmark for extracting routable pedestrian pathway graphs. The dataset includes aerial images, road graphs, road rasters, and ground truth data for multiple cities.

• Metadata: PathwayBench Croissant metadata
• License: ODbL

Installation

# create and activate the conda environment
conda create -n pathwaybench python=3.8
conda activate pathwaybench
Clone the repository and install locally as a package:

```bash
git clone https://github.com/your-username/pathways-bench.git
cd pathways-bench
pip install .

This code has been tested with Python 3.10. macOS: Sequoia

Dependencies

geopandas
shapely
osmnx
geonetworkx
pyproj

Datasets

Each set of samples in the PathwayBench dataset includes five co-registered features. The filename of each set of samples and the corresponding features are listed below:

Filename	Feature Type
xxxx_aerial.png	The aerial satellite imagery.
xxxx_road.geojson	The street (road) graph.
xxxx_road.png	The rasterized street map (with additional features).
xxxx_gt_graph.geojson	The human-validated pedestrian pathway graph.
xxxx_gt_mask.png	The rasterized human-validated pedestrian pathway graph to support semantic segmentation tasks.
xxxx_gt_color.png	The color-coded version of xxxx_gt_mask.png for visualization purposes.

Below are the links to the dataset that are currently supported by PathwayBench

City	Data
Seattle, WA	Link to dataset
Washington, D.C.	Link to dataset
Portland, OR	Link to dataset
Bellevue, WA	[Will be released soon]
Quito, Ecuador	[Will be released soon]
Sao Paulo, Brazil	[Will be released soon]
Santiago, Chile	[Will be released soon]
Valparaiso, Chile	[Will be released soon]

Benchmark

PathwayBench provides utilities for evaluating graphs by the extent to which their structural characteristics align with ground truth, as described below.

Partition test area: This step partitions the entire test area into Tessellating Intersection Polygons (TIP). Each TIP is created by assigning a point location to a road intersection, then computing the associated Voronoi polygons to tessellate the entire test area. Ground Truth GeoJSON is provided for each of the support city in PathwayBench dataset.

  from pathways_bench import PathwaysBench
  
  bench = PathwaysBench(proj='epsg:4326', debug=True)
  print(bench.version)
  output_file = bench.tessellate_area(filepath="input.geojson")
  print("Tessellated output saved at:", output_file)

Parameters

• filepath - Input GeoJSON defining area of interest (Polygon or MultiPolygon)
• output_path - Optional path to save output file
• proj - Target projection (default: epsg:26910)
• debug - Enables detailed logging when set to True