Experiential modelling of urban street: a combining approach with neural image assessment and street experiment

To design well-being and smart communities, it is important to know what street scapes/layouts are good for people experience with comfortableness, activeness, beautifulness, etc. For that purpose, walkability is one of the key performance indicators expressing the environmental quality of a street. As the first step for creating the well-being smart communities, this study attempts to evaluate the influence of street scapes/layouts by using street images taken by a volunteered geographic information application, Mapillary and a image assessment with machine learning technique. We conduct street experiments in a district in Tokyo, Japan for comparing the score of the quality of street image with the answers of questionnaire on the street. The result suggests that score of quality of images is not consistent with the street experience for people such as comfortableness, secureness, and activeness.