Soft Biometric Recognition from Comparative Crowdsourced Annotations
Soft Biometric Recognition from Comparative Crowdsourced Annotations
Soft biometrics provide cues that enable human identification from low quality video surveillance footage. This paper discusses a new crowdsourced dataset, collecting comparative soft biometric annotations from a rich set of human annotators. We now include gender as a comparative trait, and find comparative labels are more objective and obtain more accurate measurements than previous categorical labels. Using our pragmatic dataset, we perform semantic recognition by inferring relative biometric signatures. This demonstrates a practical scenario, reproducing responses from a video surveillance operator searching for an individual. The experiment is guaranteed to return the correct match in the top 7% of results with 10 comparisons, or top 13% of results using just 5 sets of subject comparisons.
Martinho-Corbishley, Daniel
6dd73e5c-9a7e-41bd-b896-fb1ea9852abb
Nixon, Mark S.
2b5b9804-5a81-462a-82e6-92ee5fa74e12
Carter, John N.
e05be2f9-991d-4476-bb50-ae91606389da
16 July 2015
Martinho-Corbishley, Daniel
6dd73e5c-9a7e-41bd-b896-fb1ea9852abb
Nixon, Mark S.
2b5b9804-5a81-462a-82e6-92ee5fa74e12
Carter, John N.
e05be2f9-991d-4476-bb50-ae91606389da
Martinho-Corbishley, Daniel, Nixon, Mark S. and Carter, John N.
(2015)
Soft Biometric Recognition from Comparative Crowdsourced Annotations.
6th International Conference on Imaging for Crime Prevention and Detection.
Record type:
Conference or Workshop Item
(Other)
Abstract
Soft biometrics provide cues that enable human identification from low quality video surveillance footage. This paper discusses a new crowdsourced dataset, collecting comparative soft biometric annotations from a rich set of human annotators. We now include gender as a comparative trait, and find comparative labels are more objective and obtain more accurate measurements than previous categorical labels. Using our pragmatic dataset, we perform semantic recognition by inferring relative biometric signatures. This demonstrates a practical scenario, reproducing responses from a video surveillance operator searching for an individual. The experiment is guaranteed to return the correct match in the top 7% of results with 10 comparisons, or top 13% of results using just 5 sets of subject comparisons.
Text
icdp-15-camera.pdf
- Accepted Manuscript
Available under License Other.
More information
Accepted/In Press date: 15 June 2015
Published date: 16 July 2015
Venue - Dates:
6th International Conference on Imaging for Crime Prevention and Detection, 2015-06-15
Organisations:
Vision, Learning and Control
Identifiers
Local EPrints ID: 380275
URI: http://eprints.soton.ac.uk/id/eprint/380275
PURE UUID: 0c8d4e66-fa97-4c92-a6d8-04c8c3d6ccdb
Catalogue record
Date deposited: 18 Aug 2015 10:40
Last modified: 15 Mar 2024 02:35
Export record
Contributors
Author:
Daniel Martinho-Corbishley
Author:
John N. Carter
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
