Presentation 2012-10-04
People Re-identification Using Local Similarity Estimation
Guanwen Zhang, Jien Kato, Yu Wang, Kenji Mase,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In this paper, we propose an efficient and robust approach for multiple-shot people re-identification. To compare the people's appearance instances of image set, we explore the multimodal property of people appearance distribution and estimate the similarity in the local neighbours of the appearance instances. An energy-based loss function is proposed to measure the distance as the similarity in feature space. This loss function favors close distances to support the people with high similarity and penalizes large distances as well as overlap between neighbours to exclude the people with low similarity. Experiments on the public datasets show significant improvements over previous reports.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Similarity Estimation / People Re-identification / Energy-based Loss Function
Paper # PRMU2012-52
Date of Issue

Conference Information
Committee PRMU
Conference Date 2012/9/27(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Pattern Recognition and Media Understanding (PRMU)
Language ENG
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) People Re-identification Using Local Similarity Estimation
Sub Title (in English)
Keyword(1) Similarity Estimation
Keyword(2) People Re-identification
Keyword(3) Energy-based Loss Function
1st Author's Name Guanwen Zhang
1st Author's Affiliation Graduate School of Information Science, Nagoya University()
2nd Author's Name Jien Kato
2nd Author's Affiliation Graduate School of Information Science, Nagoya University
3rd Author's Name Yu Wang
3rd Author's Affiliation Graduate School of Information Science, Nagoya University
4th Author's Name Kenji Mase
4th Author's Affiliation Graduate School of Information Science, Nagoya University
Date 2012-10-04
Paper # PRMU2012-52
Volume (vol) vol.112
Number (no) 225
Page pp.pp.-
#Pages 5
Date of Issue