Natural action recognition using invariant 3D motion encoding

Hadfield, S, Lebeda, K and Bowden, R Natural action recognition using invariant 3D motion encoding Proceedings of the European Conference on Computer Vision (ECCV), 8690. pp. 758-771.

	Text Hadfield_ECCV_2014pp.pdf - ["content_typename_Submitted version (pre-print)" not defined] Restricted to Repository staff only Available under License : See the attached licence file. Download (3MB)
	Text (licence) SRI_deposit_agreement.pdf Restricted to Repository staff only Available under License : See the attached licence file. Download (33kB)

Abstract

We investigate the recognition of actions "in the wild"’ using 3D motion information. The lack of control over (and knowledge of) the camera configuration, exacerbates this already challenging task, by introducing systematic projective inconsistencies between 3D motion fields, hugely increasing intra-class variance. By introducing a robust, sequence based, stereo calibration technique, we reduce these inconsistencies from fully projective to a simple similarity transform. We then introduce motion encoding techniques which provide the necessary scale invariance, along with additional invariances to changes in camera viewpoint. On the recent Hollywood 3D natural action recognition dataset, we show improvements of 40% over previous state-of-the-art techniques based on implicit motion encoding. We also demonstrate that our robust sequence calibration simplifies the task of recognising actions, leading to recognition rates 2.5 times those for the same technique without calibration. In addition, the sequence calibrations are made available.

Item Type:	Article
Divisions :	Surrey research (other units)
Authors :	Hadfield, S, Lebeda, K and Bowden, R
DOI :	10.1007/978-3-319-10605-2_49
Contributors :	Contribution Name Email ORCID http://www.loc.gov/loc.terms/relators/EDT Tuytelaars, T http://www.loc.gov/loc.terms/relators/EDT Schiele, B http://www.loc.gov/loc.terms/relators/EDT Pajdla, T http://www.loc.gov/loc.terms/relators/EDT Fleet, D
Related URLs :	http://personal.ee.surrey.ac.uk/Personal...
Depositing User :	Symplectic Elements
Date Deposited :	28 Mar 2017 10:56
Last Modified :	24 Jan 2020 12:40
URI:	http://epubs.surrey.ac.uk/id/eprint/808952

Actions (login required)

View Item

Downloads