Assessing the Pedestrian Infrastructure for Integrated Visual Walkability of Kolkata Municipal Corporation using Deep Learning based Geospatial Artificial Intelligence (Geo AI)

Abstract. Walkability, a core element of urban mobility, is indispensable for health, liveability, and sustainability. However, it continues to face challenges in the major cities of developing countries across the Global South. Adopting a case study of the metropolitan city of Kolkata, West Bengal, India, this study aims to assess the integrated visual walkability using Mapillary Street View Imagery combined with deep learning-based semantic segmentation techniques. Four factors have been considered to study the walkability: Greenery, Openness, Pavement and Crowdedness along the selected footpaths of the study area. The semantic labels are then used to quantify the selected indicators and the areas are mapped using GeoAI techniques to reveal the intra-city variations, on a normalised scale from 0- 1, where 0 indicates not walkable and 1 indicates highly walkable. The findings indicate significant discrepancies in pedestrian infrastructure, particularly within the central business districts (CBDs) of the city. These disparities are evident in inadequate footpath widths, unsafe walking spaces, and ignorance of inclusive design considerations. This shortfall in pedestrian-friendly infrastructure contributes to a less livable urban environment, impacting safety, accessibility, and overall enjoyment of the city for pedestrians. Further, the study acknowledges the potential of street view imageries and deep learning-based methods in studying urban mobility. The findings are intended to support data-driven, inclusive, and sustainable spatial planning for improved urban mobility.