About WalkSafe-AI

Project Overview

WalkSafe-AI is a data-driven pedestrian safety platform that identifies and ranks dangerous intersections using empirical Bayes analysis of crash data. The project combines PennDOT crash records, DVRPC traffic volumes, and City of Philadelphia GIS data to produce reliable risk estimates for over 16,000 intersections.

Methodology

The core ranking uses an empirical Bayes (EB) approach that combines an intersection's observed crash history with a safety performance function (SPF) fitted to similar intersections. This reduces the influence of random year-to-year variation and produces more stable, reliable risk estimates than raw crash counts alone.

Data Sources

  • -PennDOT PCDS: Pennsylvania crash data system covering pedestrian-involved crashes from 2015 through 2024.
  • -DVRPC AADT: Annual average daily traffic volume estimates from the Delaware Valley Regional Planning Commission.
  • -City of Philadelphia GIS: High Injury Network, speed camera locations, street classifications, parks, and school locations.
  • -City of Bogotá open data: street segments, transport analysis zones, crash records and built-environment attributes. Processing by Universidad de los Andes under subcontract to the WalkSafe-AI project. The detector is STRIDE, developed by the Biomedical Computer Vision group at Universidad de los Andes and released at github.com/BCV-Uniandes/STRIDE.

Team

WalkSafe-AI is developed at Drexel University as part of a research initiative on pedestrian safety and urban infrastructure analytics. This dashboard and the Bogotá analyses are supported in part by the Built Environment, Pedestrian Injuries and Deep Learning (BEPIDL) Study, NIH Fogarty International Center award K01 TW011782. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Contact

For questions, data requests, or collaboration inquiries, contact D. Alex Quistberg, Dornsife School of Public Health, Drexel University: daq26@drexel.edu.