Building Instance Classification using Social Media Images

Abstract

Understanding urbanization and planning for theupcoming changes require detailed knowledge about the placeswhere people live and work. Thus, knowing the usage of buildingsis inevitable to distinguish between residential and commercialplaces. Assessing the usage of buildings from an aerial perspectivealone is challenging and results in unresolveable ambiguities.Ascomplementary data sources, social media images taken fromground level allow access to the building fac ̧ades, as well asongoing social activities around the buildings, which are veryvaluable information while coming to accessing the buildingusages. Towards the fusion of social media images and remotesensing data for this purpose, in this work we present a methodto assess building usages from social media images taken in theirneighborhood. Using a straight forward next neighbor classifierfor mapping images to buildings and pre-trained networks fordimensionality reduction we trained a logistic regression classifierto distinguish between five different building usage classes.Applied to a study area of Los Angeles metropolitan area, USA,we obtain an average precision of 0.67. Hence, we show thatsocial media images can be a valuable additional source to remotesensing data.