Assessment of Perceived and Physical Walkability Using Street View Images and Deep Learning Technology

As neighborhood walkability has gradually become an important topic in various fields, many cities around the world are promoting eco-friendly and people-centered walking environment as a top priority urban planning. The purpose of this study is to visualize physical perceived detail analyze differences prepare alternatives for improving neighborhood’s environment. area Jeonju City, one medium-sized Korea. For evaluation walkability, 196,624 street view images were crawled 127,317 pairs training datasets constructed. After developing convolutional neural network model, scores predicted. eight indicators selected, score overall calculated by combining indicators. that, visualized, difference between them analyzed. This novel three aspects. First, we develop deep learning model that can improve accuracy using images, even small cities. Second, analyzing characteristics possibilities limitations semantic segmentation technique confirmed. Third, analyzed detail, how results our be used presented.