Efficiency Analysis of Deep Metric Learning for Visual Geolocalisation in Urban Area

Abstract

The paper is devoted to the problem of visual geolocalisation in an urban area. A method of building recognition based on convolutional neural networks (CNN) is proposed. A set of deep metric learning algorithms is analyzed to improve the recognition performance. Experimental results on ZuBuD dataset demonstrate the benefits of applying deep metric learning in the problem of visual geolocalisation.