Author(s): Ana Medina, Diana Mosquera & Francisco Alejandro Gallegos
Urban safety remains a critical challenge for social minorities, particularly in peripheral and low-income neighborhoods of Global South cities. In recent years, cities across the Global South, as elsewhere, have made major investments in efforts to promote and achieve safety, equity, and quality of life for minority communities, but these projects are nevertheless long-term efforts and require large sums of finance, data, and human resources, among other factors to be successful. In this evolving context, this study offers a novel methodology that integrates artificial intelligence (AI) models with feminist urban design principles to analyze perceptions of safety in public spaces of low-income peripheral neighborhoods in three secondary South American cities: Cuenca (Ecuador), Medellín (Colombia), and Rosario (Argentina). Using Google Street View API, we systematically collect street-level images surrounding bus stops from representative neighborhoods in each of these cities and using Real-ESRGAN super-resolution model to enhance them. The YOLO-World object detection model is then applied to these images to identify urban features, categorizing them according to five feminist urban design principles: safety, proximity, diversity, autonomy, and vitality. Our findings indicate that 82.01% of the detections correspond to safety-related features, while other principles, such as proximity (6.69%), autonomy (5.02%), vitality (4.05%) and diversity (2.23%), were identified at lower frequencies. By leveraging AI for street-level image analysis, we demonstrate a replicable method to accurately process open imagery data and qualitative analysis in such public spaces to provide urban planners and policymakers with valuable insights for developing more equitable and inclusive cities for social minorities, both in and beyond the Global South.
https://doi.org/10.35483/ACSA.AM.113.33
Volume Editors
Sara Jensen Carr & Rubén García Rubio
ISBN
978-1-944214-48-7
Study Architecture
ProPEL
