Towards Human-AI Accessibility Mapping in India: VLM-Guided Annotations and POI-Centric Analysis in Chandigarh

TL;DR

This work addresses the lack of geolocated sidewalk accessibility data in India by adapting Project Sidewalk for Chandigarh with India-specific labeling and a VLM-guided mission assistant. It provides a concrete methodology for labeling Indian street barriers, ensures context-aware guidance, and conducts a POI-centric accessibility analysis across three sectors over ~40 km and ~230 POIs, identifying 1,644 actionable locations. The contributions include interface and taxonomy localization, validation of AI-assisted guidance (mean utility 4.66), and the development of SegScore, POISecScore, and POIScore to quantify accessibility at multiple scales. The findings reveal uneven accessibility across land uses, with commercial areas typically more accessible than educational and public-service sites, highlighting practical implications for urban planning and targeted interventions in Indian cities.

Abstract

Project Sidewalk is a web-based platform that enables crowdsourcing accessibility of sidewalks at city-scale by virtually walking through city streets using Google Street View. The tool has been used in 40 cities across the world, including the US, Mexico, Chile, and Europe. In this paper, we describe adaptation efforts to enable deployment in Chandigarh, India, including modifying annotation types, provided examples, and integrating VLM-based mission guidance, which adapts instructions based on a street scene and metadata analysis. Our evaluation with 3 annotators indicates the utility of AI-mission guidance with an average score of 4.66. Using this adapted Project Sidewalk tool, we conduct a Points of Interest (POI)-centric accessibility analysis for three sectors in Chandigarh with very different land uses, residential, commercial and institutional covering about 40 km of sidewalks. Across 40 km of roads audited in three sectors and around 230 POIs, we identified 1,644 of 2,913 locations where infrastructure improvements could enhance accessibility.

Figures (5)

• Figure 1: Systematic overview of our India-adapted Project Sidewalk workflow. (a) Population extrapolation from ward-to-sector boundaries for selecting sectors for analysis. (b) Each street segment in the sector is defined by its OSM road type and start/end Street View panoramas. (c) A visual–language model (VLM) generates mission guidance from these inputs, helping annotators know what accessibility barriers to expect. (d) Annotators label barriers in the adapted Project Sidewalk interface using Indian-specific labels.
• Figure 2: Adaptations to the Project Sidewalk interface for deployment in India. (a) Overview of label-set changes, showing which categories were retained, removed, renamed, or newly introduced to reflect Indian pedestrian infrastructure. (b) Revised severity examples illustrating how the original U.S. Curb Ramp label maps to the broader Indian Curb Style category, while preserving the three-level severity scale.
• Figure 3: Representative examples of adapted label categories and respective tags in the Indian context. Images were sampled from annotated segments in Chandigarh and illustrate the diversity of visual conditions that motivated the interface redesign.
• Figure 4: Sector POI Accessibility (POISecScore) and across-sector POI accessibility (POIScore). (a) Heatmap of sectors (rows) vs. POI categories (columns), (b) Category-wise summary of POI scores aggregated across sectors.
• Figure 5: From left to right: SegScore heatmaps for sectors 12, 34, and 45 show the places of interest and the segscores for road segments. It highlights major road segments where accessibility improvements are required.