The University of Southampton
University of Southampton Institutional Repository

Mobile image parsing for visual clothing search, augmented reality mirror, and person identification

Mobile image parsing for visual clothing search, augmented reality mirror, and person identification
Mobile image parsing for visual clothing search, augmented reality mirror, and person identification
With the emergence and growing popularity of online social networks, depth sensors (such as Kinect), smart phones /tablets, wearable devices, and augmented reality (such as Google Glass and Google Cardboard), the way in which people interact with digital media has been completely transformed.

Globally, the apparel market is expected to grow at a compound annual growth rate of 5 between 2012 and 2025. Due to the huge impact for ecommerce applications, there is a growing interest in methods for clothing retrieval and outfit recommendation, especially efficient ones suitable for mobile apps. To this end, we propose a practical and efficient method for mobile visual clothing search and implement it as a smart phone app that enables the user to capture a photo of clothing of interest with their smart phone and retrieve similar clothing products that are available at nearby retailers.

Furthermore, we propose an extended method where soft biometric clothing attributes are combined with anthropometrics computed from depth data for person identification and surveillance applications. This addresses the increased terrorist threat in recent years that has driven the need for non-intrusive person identification that can operate at a distance without a subject’s knowledge or collaboration. We implement the method in a wearable mobile augmented reality application based on a smart phone with Google Cardboard in order to demonstrate how a security guard could have their vision augmented to automatically identify a suspect in their field of vision.

Lastly, we consider that a significant proportion of photos shared online and via apps are selfies and of dressed people in general. Hence, it is important both for consumers and for industry that systems are developed to understand the visual content in the vast datasets of networked content to aid management and perform smart analysis. To this end, this dissertation introduces an efficient technique to segment clothing in photos and recognize clothing attributes. We demonstrate with respect to the emerging augmented reality field by implementing an augmented reality mirror app for mobile tablet devices that can segment a user’s clothing in real-time and enable them to realistically see themselves in the mirror wearing variations of the clothing with different colours or graphics rendered. Empirical results show promising segmentation, recognition, and augmented reality performance.
Cushen, George
52f73d41-3ae0-4c11-a50a-86e782c03745
Cushen, George
52f73d41-3ae0-4c11-a50a-86e782c03745
Nixon, Mark
2b5b9804-5a81-462a-82e6-92ee5fa74e12

(2016) Mobile image parsing for visual clothing search, augmented reality mirror, and person identification. University of Southampton, Faculty of Physical Sciences and Engineering, Doctoral Thesis, 147pp.

Record type: Thesis (Doctoral)

Abstract

With the emergence and growing popularity of online social networks, depth sensors (such as Kinect), smart phones /tablets, wearable devices, and augmented reality (such as Google Glass and Google Cardboard), the way in which people interact with digital media has been completely transformed.

Globally, the apparel market is expected to grow at a compound annual growth rate of 5 between 2012 and 2025. Due to the huge impact for ecommerce applications, there is a growing interest in methods for clothing retrieval and outfit recommendation, especially efficient ones suitable for mobile apps. To this end, we propose a practical and efficient method for mobile visual clothing search and implement it as a smart phone app that enables the user to capture a photo of clothing of interest with their smart phone and retrieve similar clothing products that are available at nearby retailers.

Furthermore, we propose an extended method where soft biometric clothing attributes are combined with anthropometrics computed from depth data for person identification and surveillance applications. This addresses the increased terrorist threat in recent years that has driven the need for non-intrusive person identification that can operate at a distance without a subject’s knowledge or collaboration. We implement the method in a wearable mobile augmented reality application based on a smart phone with Google Cardboard in order to demonstrate how a security guard could have their vision augmented to automatically identify a suspect in their field of vision.

Lastly, we consider that a significant proportion of photos shared online and via apps are selfies and of dressed people in general. Hence, it is important both for consumers and for industry that systems are developed to understand the visual content in the vast datasets of networked content to aid management and perform smart analysis. To this end, this dissertation introduces an efficient technique to segment clothing in photos and recognize clothing attributes. We demonstrate with respect to the emerging augmented reality field by implementing an augmented reality mirror app for mobile tablet devices that can segment a user’s clothing in real-time and enable them to realistically see themselves in the mirror wearing variations of the clothing with different colours or graphics rendered. Empirical results show promising segmentation, recognition, and augmented reality performance.

PDF
Final PhD_Thesis George_Cushen.pdf - Other
Available under License University of Southampton Thesis Licence.
Download (19MB)

More information

Published date: February 2016
Organisations: University of Southampton, Vision, Learning and Control

Identifiers

Local EPrints ID: 400088
URI: http://eprints.soton.ac.uk/id/eprint/400088
PURE UUID: 5876e69f-a4d0-4c1a-b36f-d6cf2bd7aee3
ORCID for Mark Nixon: ORCID iD orcid.org/0000-0002-9174-5934

Catalogue record

Date deposited: 22 Sep 2016 10:30
Last modified: 06 Jun 2018 13:18

Export record

Contributors

Author: George Cushen
Thesis advisor: Mark Nixon ORCID iD

University divisions

Download statistics

Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.

View more statistics

Atom RSS 1.0 RSS 2.0

Contact ePrints Soton: eprints@soton.ac.uk

ePrints Soton supports OAI 2.0 with a base URL of http://eprints.soton.ac.uk/cgi/oai2

This repository has been built using EPrints software, developed at the University of Southampton, but available to everyone to use.

We use cookies to ensure that we give you the best experience on our website. If you continue without changing your settings, we will assume that you are happy to receive cookies on the University of Southampton website.

×