Learning class- and location-specific priors for urban semantic labeling with CNNs

Benjamin Kellenberger, Michele Volpi, Devis Tuia

Research output: Chapter in Book/Report/Conference proceedingConference paper

1 Citation (Scopus)

Abstract

This paper addresses the problem of semantic image labeling of urban remote sensing images into land cover maps. We exploit the prior knowledge that cities are composed of comparable spatial arrangements of urban objects, such as buildings. To do so, we cluster OpenStreetMap (OSM) building footprints into groups with similar local statistics, corresponding to different types of urban zones. We use the per-cluster expected building fraction to correct for over- and underrepresentations of classes predicted by a Convolutional Neural Network (CNN), using a Conditional Random Field (CRF). Results indicate a substantial improvement in both numerical and visual accuracy of the labeled maps.

Original languageEnglish
Title of host publication2017 Joint Urban Remote Sensing Event, JURSE 2017
Place of PublicationDubai
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781509058082
ISBN (Print)9781509058082
DOIs
Publication statusPublished - 10 May 2017
Externally publishedYes
Event2017 Joint Urban Remote Sensing Event, JURSE 2017 - Dubai, United Arab Emirates
Duration: 6 Mar 20178 Mar 2017

Publication series

Name2017 Joint Urban Remote Sensing Event, JURSE 2017

Conference

Conference2017 Joint Urban Remote Sensing Event, JURSE 2017
CountryUnited Arab Emirates
CityDubai
Period6/03/178/03/17

Fingerprint Dive into the research topics of 'Learning class- and location-specific priors for urban semantic labeling with CNNs'. Together they form a unique fingerprint.

Cite this