Adversarially Erased Learning for Person Re-identification by Fully Convolutional Networks

Shuangwei Liu, Yunzhou Zhang, Lin Qi, Sonya Coleman, Dermot Kerr, Shangdong Zhu

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

The generalization ability of deep person re-identification networks is subject to inadequate person data and occlusions. To relieve this dilemma, we propose a feature-level augmentation strategy, Adversarially Erased Learning Module (AELM), using two adversarial classifiers. Specifically, we utilize a classifier to identify discriminative regions and erase them to increase the variant of features. Meanwhile, we input the erased feature maps to another classifier to discover new body regions, which effectively resist occlusion of key parts. To easily perform end-to-end training for AELM, we propose a novel Identity model based on Fully Convolutional Networks (IFCN) to directly obtain body response heatmap during the forward pass by selecting corresponding class-specific feature map. Thus, the discriminative regions can be identified and erased in a convenient way. Moreover, to capture discriminative region for AELM, we present a Complementary Attention Module (CoAM) combined with channel and spatial attention to automatically focus on which feature types and positions are meaningful in the feature maps. In this paper, CoAM and AELM are cascaded into one module which is applied to the outputs of different convolutional layers to integrate mid- and high-level semantic features. Experimental results on three challenging benchmarks demonstrate the effectiveness of the proposed method.

LanguageEnglish
Title of host publication2019 International Joint Conference on Neural Networks, IJCNN 2019
Place of PublicationBudapest
Number of pages9
Volume2019-July
ISBN (Electronic)9781728119854
DOIs
Publication statusPublished - 12 Jul 2019

Fingerprint

Classifiers
Semantics

Keywords

  • Adversarially Erased Learning
  • Complementary Attention
  • Fully Convolutional Networks

Cite this

Liu, S., Zhang, Y., Qi, L., Coleman, S., Kerr, D., & Zhu, S. (2019). Adversarially Erased Learning for Person Re-identification by Fully Convolutional Networks. In 2019 International Joint Conference on Neural Networks, IJCNN 2019 (Vol. 2019-July). [8852283] Budapest. https://doi.org/10.1109/IJCNN.2019.8852283
Liu, Shuangwei ; Zhang, Yunzhou ; Qi, Lin ; Coleman, Sonya ; Kerr, Dermot ; Zhu, Shangdong. / Adversarially Erased Learning for Person Re-identification by Fully Convolutional Networks. 2019 International Joint Conference on Neural Networks, IJCNN 2019. Vol. 2019-July Budapest, 2019.
@inproceedings{bf9d4f427d6948aa8b4bab4490a5371f,
title = "Adversarially Erased Learning for Person Re-identification by Fully Convolutional Networks",
abstract = "The generalization ability of deep person re-identification networks is subject to inadequate person data and occlusions. To relieve this dilemma, we propose a feature-level augmentation strategy, Adversarially Erased Learning Module (AELM), using two adversarial classifiers. Specifically, we utilize a classifier to identify discriminative regions and erase them to increase the variant of features. Meanwhile, we input the erased feature maps to another classifier to discover new body regions, which effectively resist occlusion of key parts. To easily perform end-to-end training for AELM, we propose a novel Identity model based on Fully Convolutional Networks (IFCN) to directly obtain body response heatmap during the forward pass by selecting corresponding class-specific feature map. Thus, the discriminative regions can be identified and erased in a convenient way. Moreover, to capture discriminative region for AELM, we present a Complementary Attention Module (CoAM) combined with channel and spatial attention to automatically focus on which feature types and positions are meaningful in the feature maps. In this paper, CoAM and AELM are cascaded into one module which is applied to the outputs of different convolutional layers to integrate mid- and high-level semantic features. Experimental results on three challenging benchmarks demonstrate the effectiveness of the proposed method.",
keywords = "Adversarially Erased Learning, Complementary Attention, Fully Convolutional Networks",
author = "Shuangwei Liu and Yunzhou Zhang and Lin Qi and Sonya Coleman and Dermot Kerr and Shangdong Zhu",
year = "2019",
month = "7",
day = "12",
doi = "10.1109/IJCNN.2019.8852283",
language = "English",
volume = "2019-July",
booktitle = "2019 International Joint Conference on Neural Networks, IJCNN 2019",

}

Liu, S, Zhang, Y, Qi, L, Coleman, S, Kerr, D & Zhu, S 2019, Adversarially Erased Learning for Person Re-identification by Fully Convolutional Networks. in 2019 International Joint Conference on Neural Networks, IJCNN 2019. vol. 2019-July, 8852283, Budapest. https://doi.org/10.1109/IJCNN.2019.8852283

Adversarially Erased Learning for Person Re-identification by Fully Convolutional Networks. / Liu, Shuangwei; Zhang, Yunzhou; Qi, Lin; Coleman, Sonya; Kerr, Dermot; Zhu, Shangdong.

2019 International Joint Conference on Neural Networks, IJCNN 2019. Vol. 2019-July Budapest, 2019. 8852283.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Adversarially Erased Learning for Person Re-identification by Fully Convolutional Networks

AU - Liu, Shuangwei

AU - Zhang, Yunzhou

AU - Qi, Lin

AU - Coleman, Sonya

AU - Kerr, Dermot

AU - Zhu, Shangdong

PY - 2019/7/12

Y1 - 2019/7/12

N2 - The generalization ability of deep person re-identification networks is subject to inadequate person data and occlusions. To relieve this dilemma, we propose a feature-level augmentation strategy, Adversarially Erased Learning Module (AELM), using two adversarial classifiers. Specifically, we utilize a classifier to identify discriminative regions and erase them to increase the variant of features. Meanwhile, we input the erased feature maps to another classifier to discover new body regions, which effectively resist occlusion of key parts. To easily perform end-to-end training for AELM, we propose a novel Identity model based on Fully Convolutional Networks (IFCN) to directly obtain body response heatmap during the forward pass by selecting corresponding class-specific feature map. Thus, the discriminative regions can be identified and erased in a convenient way. Moreover, to capture discriminative region for AELM, we present a Complementary Attention Module (CoAM) combined with channel and spatial attention to automatically focus on which feature types and positions are meaningful in the feature maps. In this paper, CoAM and AELM are cascaded into one module which is applied to the outputs of different convolutional layers to integrate mid- and high-level semantic features. Experimental results on three challenging benchmarks demonstrate the effectiveness of the proposed method.

AB - The generalization ability of deep person re-identification networks is subject to inadequate person data and occlusions. To relieve this dilemma, we propose a feature-level augmentation strategy, Adversarially Erased Learning Module (AELM), using two adversarial classifiers. Specifically, we utilize a classifier to identify discriminative regions and erase them to increase the variant of features. Meanwhile, we input the erased feature maps to another classifier to discover new body regions, which effectively resist occlusion of key parts. To easily perform end-to-end training for AELM, we propose a novel Identity model based on Fully Convolutional Networks (IFCN) to directly obtain body response heatmap during the forward pass by selecting corresponding class-specific feature map. Thus, the discriminative regions can be identified and erased in a convenient way. Moreover, to capture discriminative region for AELM, we present a Complementary Attention Module (CoAM) combined with channel and spatial attention to automatically focus on which feature types and positions are meaningful in the feature maps. In this paper, CoAM and AELM are cascaded into one module which is applied to the outputs of different convolutional layers to integrate mid- and high-level semantic features. Experimental results on three challenging benchmarks demonstrate the effectiveness of the proposed method.

KW - Adversarially Erased Learning

KW - Complementary Attention

KW - Fully Convolutional Networks

UR - http://www.scopus.com/inward/record.url?scp=85073246923&partnerID=8YFLogxK

U2 - 10.1109/IJCNN.2019.8852283

DO - 10.1109/IJCNN.2019.8852283

M3 - Conference contribution

VL - 2019-July

BT - 2019 International Joint Conference on Neural Networks, IJCNN 2019

CY - Budapest

ER -

Liu S, Zhang Y, Qi L, Coleman S, Kerr D, Zhu S. Adversarially Erased Learning for Person Re-identification by Fully Convolutional Networks. In 2019 International Joint Conference on Neural Networks, IJCNN 2019. Vol. 2019-July. Budapest. 2019. 8852283 https://doi.org/10.1109/IJCNN.2019.8852283