When providing directions to a place, web and mobile mapping services are allable to suggest the shortest route. The goal of this work is to automaticallysuggest routes that are not only short but also emotionally pleasant. Toquantify the extent to which urban locations are pleasant, we use data from acrowd-sourcing platform that shows two street scenes in London (out ofhundreds), and a user votes on which one looks more beautiful, quiet, andhappy. We consider votes from more than 3.3K individuals and translate theminto quantitative measures of location perceptions. We arrange those locationsinto a graph upon which we learn pleasant routes. Based on a quantitativevalidation, we find that, compared to the shortest routes, the recommended onesadd just a few extra walking minutes and are indeed perceived to be morebeautiful, quiet, and happy. To test the generality of our approach, weconsider Flickr metadata of more than 3.7M pictures in London and 1.3M inBoston, compute proxies for the crowdsourced beauty dimension (the one forwhich we have collected the most votes), and evaluate those proxies with 30participants in London and 54 in Boston. These participants have not only ratedour recommendations but have also carefully motivated their choices, providinginsights for future work.
展开▼