This is the companion page for our WWW poster submission, titled “Predicting Urban Attractiveness Using Convolutional Neural Networks with Crowdsourced Street-Level Imagery Data“.

Abstract

Understanding how people perceive attractive or unattractive places in cities is vitally important to urban planning and policy making. Given the subjective nature of human perception and the ambiguous character of attractiveness as an attribute of urban places, it is challenging to assess their perceived attractiveness. This paper introduces a scalable approach to predicting urban attractiveness that employs a Convolutional Neural Network (CNN)-based model on crowdsourced street-level imagery data. Extensive validation shows that our approach significantly outperforms baseline methods for predicting urban attractiveness and also enables the identification of meaningful visual patterns that contribute positively or negatively to the attractiveness of an urban place. Understanding how people perceive attractive or unattractive places in cities is vitally important to urban planning and policy making. Given the subjective nature of human perception and the ambiguous character of attractiveness as an attribute of urban places, it is challenging to assess their perceived attractiveness. This paper introduces a scalable approach to predicting urban attractiveness that employs a Convolutional Neural Network (CNN)-based model on crowdsourced street-level imagery data. Extensive validation shows that our approach significantly outperforms baseline methods for predicting urban attractiveness and also enables the identification of meaningful visual patterns that contribute positively or negatively to the attractiveness of an urban place.