Self-Channel Attention Weighted Part for Person Re-Identification
Summary :
Part based models have been proved to be beneficial for person re-identification (Re-ID) in recent years. Existing models usually use fixed horizontal stripes or rely on human keypoints to get each part, which is not consistent with the human visual mechanism. In this paper, we propose a Self-Channel Attention Weighted Part model (SCAWP) for Re-ID. In SCAWP, we first learn a feature map from ResNet50 and use 1x1 convolution to reduce the dimension of this feature map, which could aggregate the channel information. Then, we learn the weight map of attention within each channel and multiply it with the feature map to get each part. Finally, each part is used for a special identification task to build the whole model. To verify the performance of SCAWP, we conduct experiment on three benchmark datasets, including CUHK03-NP, Market-1501 and DukeMTMC-ReID. SCAWP achieves rank-1/mAP accuracy of 70.4%/68.3%, 94.6%/86.4% and 87.6%/76.8% on three datasets respectively.
- Publication
- IEICE TRANSACTIONS on Fundamentals Vol.E104-A No.3 pp.665-670
- Publication Date
- 2021/03/01
- Publicized
- 2020/09/01
- Online ISSN
- 1745-1337
- DOI
- 10.1587/transfun.2020EAL2059
- Type of Manuscript
- LETTER
- Category
- Image
Authors
Lin DU
Army Engineering University of PLA
Chang TIAN
Army Engineering University of PLA
Mingyong ZENG
the Jiangnan Institute of Computing Technology
Jiabao WANG
Army Engineering University of PLA
Shanshan JIAO
Army Engineering University of PLA
Qing SHEN
Army Engineering University of PLA
Wei BAI
Army Engineering University of PLA
Aihong LU
Army Engineering University of PLA
Keyword
Latest Issue
Copyrights notice of machine-translated contents
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Cite this
Copy
Lin DU, Chang TIAN, Mingyong ZENG, Jiabao WANG, Shanshan JIAO, Qing SHEN, Wei BAI, Aihong LU, "Self-Channel Attention Weighted Part for Person Re-Identification" in IEICE TRANSACTIONS on Fundamentals,
vol. E104-A, no. 3, pp. 665-670, March 2021, doi: 10.1587/transfun.2020EAL2059.
Abstract: Part based models have been proved to be beneficial for person re-identification (Re-ID) in recent years. Existing models usually use fixed horizontal stripes or rely on human keypoints to get each part, which is not consistent with the human visual mechanism. In this paper, we propose a Self-Channel Attention Weighted Part model (SCAWP) for Re-ID. In SCAWP, we first learn a feature map from ResNet50 and use 1x1 convolution to reduce the dimension of this feature map, which could aggregate the channel information. Then, we learn the weight map of attention within each channel and multiply it with the feature map to get each part. Finally, each part is used for a special identification task to build the whole model. To verify the performance of SCAWP, we conduct experiment on three benchmark datasets, including CUHK03-NP, Market-1501 and DukeMTMC-ReID. SCAWP achieves rank-1/mAP accuracy of 70.4%/68.3%, 94.6%/86.4% and 87.6%/76.8% on three datasets respectively.
URL: https://global.ieice.org/en_transactions/fundamentals/10.1587/transfun.2020EAL2059/_p
Copy
@ARTICLE{e104-a_3_665,
author={Lin DU, Chang TIAN, Mingyong ZENG, Jiabao WANG, Shanshan JIAO, Qing SHEN, Wei BAI, Aihong LU, },
journal={IEICE TRANSACTIONS on Fundamentals},
title={Self-Channel Attention Weighted Part for Person Re-Identification},
year={2021},
volume={E104-A},
number={3},
pages={665-670},
abstract={Part based models have been proved to be beneficial for person re-identification (Re-ID) in recent years. Existing models usually use fixed horizontal stripes or rely on human keypoints to get each part, which is not consistent with the human visual mechanism. In this paper, we propose a Self-Channel Attention Weighted Part model (SCAWP) for Re-ID. In SCAWP, we first learn a feature map from ResNet50 and use 1x1 convolution to reduce the dimension of this feature map, which could aggregate the channel information. Then, we learn the weight map of attention within each channel and multiply it with the feature map to get each part. Finally, each part is used for a special identification task to build the whole model. To verify the performance of SCAWP, we conduct experiment on three benchmark datasets, including CUHK03-NP, Market-1501 and DukeMTMC-ReID. SCAWP achieves rank-1/mAP accuracy of 70.4%/68.3%, 94.6%/86.4% and 87.6%/76.8% on three datasets respectively.},
keywords={},
doi={10.1587/transfun.2020EAL2059},
ISSN={1745-1337},
month={March},}
Copy
TY - JOUR
TI - Self-Channel Attention Weighted Part for Person Re-Identification
T2 - IEICE TRANSACTIONS on Fundamentals
SP - 665
EP - 670
AU - Lin DU
AU - Chang TIAN
AU - Mingyong ZENG
AU - Jiabao WANG
AU - Shanshan JIAO
AU - Qing SHEN
AU - Wei BAI
AU - Aihong LU
PY - 2021
DO - 10.1587/transfun.2020EAL2059
JO - IEICE TRANSACTIONS on Fundamentals
SN - 1745-1337
VL - E104-A
IS - 3
JA - IEICE TRANSACTIONS on Fundamentals
Y1 - March 2021
AB - Part based models have been proved to be beneficial for person re-identification (Re-ID) in recent years. Existing models usually use fixed horizontal stripes or rely on human keypoints to get each part, which is not consistent with the human visual mechanism. In this paper, we propose a Self-Channel Attention Weighted Part model (SCAWP) for Re-ID. In SCAWP, we first learn a feature map from ResNet50 and use 1x1 convolution to reduce the dimension of this feature map, which could aggregate the channel information. Then, we learn the weight map of attention within each channel and multiply it with the feature map to get each part. Finally, each part is used for a special identification task to build the whole model. To verify the performance of SCAWP, we conduct experiment on three benchmark datasets, including CUHK03-NP, Market-1501 and DukeMTMC-ReID. SCAWP achieves rank-1/mAP accuracy of 70.4%/68.3%, 94.6%/86.4% and 87.6%/76.8% on three datasets respectively.
ER -
English
