Automatic Building Age Prediction from Street View Images

Abstract

Building age is a key factor for building energy efficiency, valuation of real estate objects and urban planning, while previous research has been limited by the available building age data and efficient ways to estimate building age information. This paper presents an automated workflow for estimating building age from street view images. A building age dataset consisting of street view images that are labeled with the date of construction is created for Amsterdam. We designed a deep convolutional neural network for the estimation of building age and achieved a total accuracy of 81%. This research utilizes publicly available data, street view images, and construction dates of buildings, to perform the estimation of building age with an automated manner.