引入語義部位約束的行人再識別

陳瑩; 陳巧媛

doi:10.11999/JEIT190954

引入語義部位約束的行人再識別

doi: 10.11999/JEIT190954

陳瑩^,,
陳巧媛

江南大學(xué)輕工過程先進(jìn)控制教育部重點(diǎn)實(shí)驗(yàn)室無錫 214122

基金項(xiàng)目: 國家自然科學(xué)基金(61573168)，江蘇省六大人才高峰資助項(xiàng)目(2015-WLW-004)

詳細(xì)信息

作者簡介:
陳瑩：女，1976年生，教授，博士生導(dǎo)師，主要研究方向?yàn)樾畔⑷诤?、模式識別等

陳巧媛：女，1995年生，碩士生，研究方向?yàn)樾腥嗽僮R別

通訊作者:
陳瑩　chenying@jiangnan.edu.cn

中圖分類號: TN911.73; TP391
計量
- 文章訪問數(shù): 1361
- HTML全文瀏覽量: 639
- PDF下載量: 81
- 被引次數(shù): 0
出版歷程
- 收稿日期: 2019-11-27
- 修回日期: 2020-06-04
- 網(wǎng)絡(luò)出版日期: 2020-07-28
- 刊出日期: 2020-12-08

Semantic Part Constraint for Person Re-identification

Ying CHEN^,,
Qiaoyuan CHEN

Key Laboratory of Advanced Control Education in Light Industry Process, Jiangnan University, Wuxi 214122, China

Funds: The National Natural Science Foundation of China (61573168), The Six Talent Summit Project Talents of Jiangsu Province (2015-WLW-004)

摘要

摘要: 為減輕行人圖片中的背景干擾，使網(wǎng)絡(luò)著重于行人前景并且提高前景中人體部位的利用率，該文提出引入語義部位約束(SPC)的行人再識別網(wǎng)絡(luò)。在訓(xùn)練階段，首先將行人圖片同時輸入主干網(wǎng)絡(luò)和語義部位分割網(wǎng)絡(luò)，分別得到行人特征圖和部位分割圖；然后，將部位分割圖與行人特征圖融合，得到語義部位特征；接著，對行人特征圖進(jìn)行池化得到全局特征；最后，同時使用身份約束和語義部位約束訓(xùn)練網(wǎng)絡(luò)。在測試階段，由于語義部位約束使得全局特征擁有部位信息，因此測試時僅使用主干網(wǎng)絡(luò)提取行人的全局信息即可。在大規(guī)模公開數(shù)據(jù)集上的實(shí)驗(yàn)結(jié)果表明，語義部位約束能有效使得網(wǎng)絡(luò)提高辨別行人身份的能力并且縮減推斷網(wǎng)絡(luò)的計算花費(fèi)。與現(xiàn)有方法比較，該文網(wǎng)絡(luò)能更好地抵抗背景干擾，提高行人再識別性能。
- 行人再識別 /
- 人體語義分割 /
- 語義部位約束
Abstract: In order to alleviate the background clutter in pedestrian images, and make the network focus on pedestrian foreground to improve the utilization of human body parts in the foreground. In this paper, a person re-identification network is proposed that introduces Semantic Part Constraint(SPC). Firstly, the pedestrian image is input into the backbone network and the semantic part segmentation network at the same time, and the pedestrian feature map and the part segmentation label are obtained respectively. Secondly, the part segmentation label and the pedestrian feature maps are merged to obtain the semantic part feature. Thirdly, the pedestrian feature map is obtained and the global average pooling is used to gain global features. Finally, the network is trained using both identity constraint and semantic part constraint. Since the semantic part constraint makes the global features obtain the part information, only the backbone network can be used to extract the features of the pedestrian during the test. Experiments on large-scale datasets show that semantic part constraints can effectively make the network improve the ability to identify pedestrians and reduce the computational cost of inferring networks. Compared with the state of art, the proposed network can better resist background clutter and improve person re-identification performance.
- Person re-identification /
- Human semantic segmentation /
- Semantic Part Constraint (SPC)

HTML全文

圖 1 本文網(wǎng)絡(luò)結(jié)構(gòu)圖

下載: 全尺寸圖片幻燈片

圖 2 語義部位標(biāo)簽示例

下載: 全尺寸圖片幻燈片

圖 3 $\lambda $的取值對應(yīng)Rank-1精度

下載: 全尺寸圖片幻燈片

圖 4 行人檢索結(jié)果排序圖

下載: 全尺寸圖片幻燈片

表 1 在Market-1501數(shù)據(jù)集上的對比實(shí)驗(yàn)(%)

實(shí)驗(yàn)編號	行人特征	網(wǎng)絡(luò)約束	Rank-1	Rank-5	Rank-10	mAP
1	${{{f}}_{\rm{g}}}$	${L_{{\rm{id}}}}$	92.0	96.9	98.2	80.4
2	${{{C}}_f}$	${L_{{\rm{id}}}}$	92.7	97.5	98.6	80.6
3	${{{f}}_{\rm{g}}}$	${L_{{\rm{id}}}} + {L_{{\rm{sp}}}}$	93.6	97.6	98.7	83.6

下載: 導(dǎo)出CSV

表 2 不同網(wǎng)絡(luò)測試時長對比(ms)

方法	批次特征提取耗時
復(fù)現(xiàn)SPReID	82.87
本文網(wǎng)絡(luò)	9.45

下載: 導(dǎo)出CSV

表 3 不同方法在兩個數(shù)據(jù)集上的性能比較(%)

方法	Market-1501		DukeMTMC-reID
方法	Rank-1	mAP	Rank-1	mAP
VIM^[11]	79.5	59.9	68.9	49.3
SVDNet^[12]	82.3	62.1	76.7	56.8
APR^[3]	84.3	64.7	70.7	51.2
FMN^[13]	86.0	67.1	74.5	56.9
PSE^[14]	87.7	69.0	79.8	62.0
PN-GAN^[15]	89.4	72.6	73.6	53.2
CamStyle^[16]	89.5	71.6	78.3	57.6
HA-CNN^[17]	91.2	75.7	80.5	63.8
Part-Aligned^[4]	91.7	79.6	84.4	69.3
SPReID^[5]	92.5	81.3	84.4	71.0
AHR^[18]	93.1	76.2	81.7	65.9
本文方法	93.6	83.6	85.4	71.3

下載: 導(dǎo)出CSV

參考文獻(xiàn)(18)

LIAO Shengcai, HU Yang, ZHU Xiangyu, et al. Person re-identification by local maximal occurrence representation and metric learning[C]. 2015 IEEE Conference on Computer Vision and Pattern Recognition, Boston, America, 2015: 2197–2206. doi: 10.1109/CVPR.2015.7298832.

陳瑩, 許瀟月. 基于雙向參考集矩陣度量學(xué)習(xí)的行人再識別[J]. 電子與信息學(xué)報, 2020, 42(2): 394–402. doi: 10.11999/JEIT190159

CHEN Ying and XU Xiaoyue. Matrix metric learning for person re-identification based on bidirectional reference set[J]. Journal of Electronics &Information Technology, 2020, 42(2): 394–402. doi: 10.11999/JEIT190159

LIN Yutian, ZHENG Liang, ZHENG Zhedong, et al. Improving person re-identification by attribute and identity Learning[J]. Pattern Recognition, 2019, 95: 151–161. doi: 10.1016/j.patcog.2019.06.006

SUH Y, WANG Jingdong, TANG Siyu, et al. Part-aligned bilinear representations for person re-identification[C]. The 15th European Conference on Computer Vision, Munich, Germany, 2018: 402–419. doi: 10.1007/978-3-030-01264-9_25.

KALAYEH M M, BASARAN E, GOKMEN M, et al. Human semantic parsing for person re-identification[C]. The IEEE/CVF Conference on Computer Vision and Pattern Recognition, Salt Lake City, America, 2018: 1062–1071. doi: 10.1109/CVPR.2018.00117.

HE Kaiming, ZHANG Xiangyu, REN Shaoqing, et al. Deep residual learning for image recognition[C]. 2016 IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, America, 2016: 770–778. doi: 10.1109/CVPR.2016.90.

FU Jun, LIU Jing, TIAN Haijie, et al. Dual attention network for scene segmentation[C]. The 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Long Beach, America, 2019: 3146–3154.

GüLER R A, NEVEROVA N, and KOKKINOS I. DensePose: Dense human pose estimation in the wild[C]. 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Salt Lake City, America, 2018: 7297–7306. doi: 10.1109/CVPR.2018.00762.

ZHENG Liang, SHEN Liyue, TIAN Lu, et al. Scalable person re-identification: A benchmark[C]. 2015 IEEE International Conference on Computer Vision, Santiago, Chile, 2015: 1116–1124. doi: 10.1109/ICCV.2015.133.

ZHENG Zhedong, ZHENG Liang, and YANG Yi. Unlabeled samples generated by GAN improve the person re-identification baseline in vitro[C]. 2017 IEEE International Conference on Computer Vision, Venice, Italy, 2017: 3754–3762. doi: 10.1109/ICCV.2017.405.

ZHENG Zhedong, ZHENG Liang, and YANG Yi. A discriminatively learned CNN embedding for person reidentification[J]. ACM Transactions on Multimedia Computing, Communications, and Applications, 2018, 14(1): 13. doi: 10.1145/3159171

SUN Yifan, ZHENG Liang, DENG Weijian, et al. SVDNet for pedestrian retrieval[C]. 2017 IEEE International Conference on Computer Vision, Venice, Italy, 2017: 3800–3808. doi: 10.1109/ICCV.2017.410.

DING Guodong, KHAN S, TANG Zhenmin, et al. Feature mask network for person re-identification[J]. Pattern Recognition Letters, 2020, 137: 91–98. doi: 10.1016/j.patrec.2019.02.015

SARFRAZ M S, SCHUMANN A, EBERLE A, et al. A pose-sensitive embedding for person re-identification with expanded cross neighborhood re-ranking[C]. The IEEE/CVF Conference on Computer Vision and Pattern Recognition, Salt Lake City, America, 2018: 420–429. doi: 10.1109/CVPR.2018.00051.

QIAN Xuelin, FU Yanwei, XIANG Tao, et al. Pose-normalized image generation for person re-identification[C]. The 15th European Conference on Computer Vision, Munich, Germany, 2018: 650–667. doi: 10.1007/978-3-030-01240-3_40.

ZHONG Zhun, ZHENG Liang, ZHENG Zhedong, et al. CamStyle: A novel data augmentation method for person re-identification[J]. IEEE Transactions on Image Processing, 2019, 28(3): 1176–1190. doi: 10.1109/TIP.2018.2874313

LI Wei, ZHU Xiatian, and GONG Shaogang. Harmonious attention network for person re-identification[C]. The 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Salt Lake City, America, 2018: 2285–2294. doi: 10.1109/CVPR.2018.00243.

陳鴻昶, 吳彥丞, 李邵梅, 等. 基于行人屬性分級識別的行人再識別[J]. 電子與信息學(xué)報, 2019, 41(9): 2239–2246. doi: 10.11999/JEIT180740

CHEN Hongchang, WU Yancheng, LI Shaomei, et al. Person re-identification based on attribute hierarchy recognition[J]. Journal of Electronics &Information Technology, 2019, 41(9): 2239–2246. doi: 10.11999/JEIT180740

相關(guān)文章

施引文獻(xiàn)

資源附件(0)

訪問統(tǒng)計