What You See Is What You Get: Data-Informed Workflow in Design for Architecture and Urbanism

This research leverages computer vision with statistical, computational, and urban design techniques to reveal insightful and helpful relations between spatial features and patterns of spatial use. Three case studies are developed in two locations, in Madrid and Pittsburgh, with distinct spatial and utilization characteristics for covering diverse conditions through two types of experiments. The first one develops a bivariate analysis between parameters from data that describe the layout and use of each location for identifying correlations between spatial features and utilization patterns. The second experiment uses machine-learning techniques for spatial clustering supported by matching temporal signatures of the detected occupations by the developed computer vision algorithms.

The proposed design framework yields new types of data about people’s interactions with their urban built environment for generating new knowledge to inform design and policy decisions more effectively and expanding the capabilities of practitioners for focusing on exploring new scenarios. This thesis advocates for expanding design through data.