Title

ArkiCity: Analysing the Object Detection Performance of Cloud-Based Image Processing Services Using Crowdsourced Data

Publication Name

Green Energy and Technology

Abstract

ArkiCity is an innovative smartphone application that enables citizens to share local knowledge and ideas about their city by taking a photo, augmenting that photo with a set of 2D and/or 3D objects and submitting both images to a backend storage service. As an effective crowdsourcing tool for smart cities, ArkiCity advocates for a more inclusive approach to urban design and bridging the gap between citizens, planners and decision-makers. ArkiCity has been trialled in ten local city councils in Australia and Denmark and has received near 1,000 end-user submissions. In this chapter, the utilisation of such crowdsourced data for training object detection models is proposed, and the performance of a cloud-based computer vision service, in terms of its accuracy in identifying the cutout objects used in these submissions, is analysed.

Open Access Status

This publication is not available as open access

First Page

381

Last Page

394

Share

COinS
 

Link to publisher version (DOI)

http://dx.doi.org/10.1007/978-3-030-96429-0_18