Towards a Computational Approach to Quantify Human Experience in Urban Design: A Data Collection Platform

Design features that form urban settings such as greenery, height of buildings, and variation in the building façade (materials, color, and proportion) are known to have effects on how people experience environments. As the urban population grows and shifts to urban settings for living (e.g., 82% of people live in cities in the US), understanding the impact of urban environments on human experience becomes more essential. Previous studies to capture human experience in urban settings have been limited due to the labor-intensive and manual process of data collection (i.e., field surveys). Due to limited quantified data on urban design features, previous methodologies were constrained to a few neighborhoods, hence lacked generalizability across regions. Advancements in technologies such as GIS, computer vision, and data-driven methodologies and accessibility to large image sets on urban settings provide opportunities to eliminate the labor-intensive process of data collection. With the help of technologies, it is possible to quantify how people experience cities. This study leverages such advancements and aims to develop an automated approach for quantifying human experience toward built environments regarding their restorative impact on citizens. Towards this aim, within the context of this paper, we provide the details of a web-based crowdsourcing platform developed for the data collection at urban scale. We combine Geographic Information System (GIS), Google Street View (GSV), and JavaScript libraries to build the platform to capture the responses of participants on the restorative impact of the environments displayed to them as images. Based on the geolocation information obtained from GIS, we collected high resolution 360? GSV images within New York City (NYC) and used them to collect responses of citizens on structured questions tailored to the scope of the study. The crowdsourcing platform enables participants to evaluate the overall restorative impact of environments given in a 360? image, and to specify areas and design features influencing their evaluation. To quantify the influential design features on responses, we use semantic segmentation, and perform statistical analysis on the dataset to examine the impact of each urban design feature on the overall restorative impact. The approach will be presented for researchers to integrate GIS, Google API, and libraries to pull massive urban data for research study necessitating a good representation of the built environment as inputs. The outcomes will guide practitioners in urban redevelopment projects about urban design features that are influential.