Mapping Public Space Micro-Occupations: Drone-Driven Predictions of Spatial Behaviors in Carapungo, Quito

Abstract

This study evaluates the spatial behavior of an intermodal transportation hub in Carapungo, one of the densest neighborhoods in Quito, Ecuador. This public infrastructure is deficient and lacks adequate equipment for the people who use, occupy, and transit within and around it, as well as for the numerous activities that occur, particularly at Carapungo’s Entry Park. Traditional methods for analyzing urban dynamics and land use are typically rigid and fail to grasp the complex and nonlinear nature of public spaces, especially in informal Global South cities. However, recent advancements in Artificial Intelligence and Machine Learning, combined with aerial drone videos, have enabled the modeling and prediction of urban dynamics beyond state regulations and formal planning. In this context, we developed a model using Computer Vision Technology and the YOLOv5 algorithm, incorporating Deep Learning training. The objective is twofold: firstly, to detect people, their movement and speed; and secondly, to produce “Occupancy” and “Count & Speed” cartographies that highlight commuters’ spatial patterns. These situated cartographies provide valuable insights into urban design, mobility, and interaction within a conflicted public space’s-built environment. The generated data offer planners and policymakers quantitative spatial information to consider local practices and dynamics in urban planning, particularly in situations of informality and insufficient urban infrastructure.