基于无参注意力和联合损失的行人重识别

程述立; 汪烈军; 王有丹

doi:10.13568/j.cnki.651094.651316.2022.07.09.0001

您当前的位置：

首页 >

文章列表页 >

基于无参注意力和联合损失的行人重识别

信息科学与技术 | 更新时间：2026-01-21

- 基于无参注意力和联合损失的行人重识别
- 基于无参注意力和联合损失的行人重识别
- 新疆大学学报（自然科学版中英文） 2023年40卷第2期页码：202-209
- 作者机构：
  
  1. 新疆大学信息科学与工程学院
  2. 新疆大学数学与系统科学学院
- 作者简介：
- 基金信息：
  
  国家自然科学基金“音视频多模态协同的异常事件鲁棒性检测关键技术研究”（U1903213）;新疆维吾尔自治区自然科学基金“基于深度学习的遥感变化检测方法研究”（2022D01C82）
- DOI：10.13568/j.cnki.651094.651316.2022.07.09.0001
  中图分类号： TP391.41
- 纸质出版：2023
- 稿件说明：
移动端阅览
[1]程述立,汪烈军,王有丹.基于无参注意力和联合损失的行人重识别[J].新疆大学学报(自然科学版)(中英文),2023,40(02):202-209.
[1]程述立,汪烈军,王有丹.基于无参注意力和联合损失的行人重识别[J].新疆大学学报(自然科学版)(中英文),2023,40(02):202-209. DOI： 10.13568/j.cnki.651094.651316.2022.07.09.0001.

DOI：10.13568/j.cnki.651094.651316.2022.07.09.0001.

摘要

现阶段行人重识别一般只考虑二维特征，将各个特征点统一处理，存在特征提取不足的问题，故提出基于无参注意力的行人重识别（PFNet）来解决上述问题．该模型在Res Net-50网络上进行改进，分别在第一个残差块和第三个残差块后引入无参注意力机制，该注意力机制能根据图片本身特点赋予各特征点合适的权重，可以保留更丰富的信息特征且不会引入额外参数．接着使用自适应平均池化层保留主要特征且捕捉特定域的判别特征，然后使用ID损失、三元组损失和自适应加权排序损失的联合损失函数来训练模型．算法在Market-1501、DukeMTMC-reID和CUHK03三个主流的行人重识别数据集上的首位命中率分别达到95.5%、90.9%和84.3%，平均精度均值分别达到89.6%、81.6%和82.0%．实验结果表明，使用注意力和联合损失函数的策略可以提高模型精度．

Abstract

At present

pedestrian re-recognition usually only considers two-dimensional features and deals with each feature point uniformly

which has the problem of insufficient feature extraction. In this paper

pedestrian re-recognition based on parameter-free attention(PFNet) is proposed to solve the above problems. The model is improved on Res Net-50 network

and the non-parametric attention mechanism is introduced after the first residual block and the third residual block respectively. The attention mechanism can assign appropriate weight to each feature point according to the characteristics of the image itself

and can retain richer information features without introducing additional parameters. Then

the adaptive average pooling layer is used to retain the main features and capture the discriminant features of a specific domain. Then

the model is trained with a combined loss function of ID loss

triplet loss and adaptive weighted ranking loss. The algorithm achieves 95.5%

90.9% and 84.3%

respectively; On the three mainstream pedestrian re-recognition datasets of Market-1501

DukeMTMC-reID and CUHK03

and the average accuracy is 89.6%

81.6% and 82.0%

respectively. Experimental results show that the strategy using attention and joint loss function can improve the model accuracy.

关键词

Keywords

references

YE M, SHEN J, LIN G, et al. Deep learning for person re-identification:a survey and outlook[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 44(6):2872-2893.

HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition[C]//2016 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). Las Vegas:IEEE, 2016.

HUANG G, LIU Z, LAURENS V, et al. Densely connected convolutional networks[C]//2017 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). Honolulu:IEEE, 2017.

ALMAHAIRI A, BALLAS N, COOIJMANS T, et al. Dynamic capacity networks[C]//The 33rd International Conference on Machine Learning(ICML). New York:ACM, 2016.

JIE H, LI S, GANG S, et al. Squeeze-and-excitation networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2020, 42(8):2011-2023.

WOO S, PARK J, LEE J Y, et al. CBAM:convolutional block attention module[C]//The European Conference on Computer Vision(ECCV). Munchen:Springer, 2018.

PARK J, WOO S, LEE J Y, et al. BAM:bottleneck attention module[C]//British Machine Vision Conference(BMVC). Newcastle:Springer, 2018.

TAY C P, YAP K H, ROY S. AANet:attribute attention network for person re-identifications[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). Long Beach:IEEE, 2019.

ZHANG Z, LAN C, ZENG W, et al. Relation-aware global attention for person re-identification[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). Seattle:IEEE, 2020.

YANG B, SHAN Y, PENG R, et al. A feature extraction method for person re-identification based on a two-branch CNN[J].Multimedia Tools and Applications, 2022, 81(1):39169-39184.

WEI L, ZHANG S, WEN G, et al. Person transfer GAN to bridge domain gap for person re-identification[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). Salt Lake City:IEEE, 2018.

DAI Z, CHEN M, GU X, et al. Batch dropblock network for person re-identification and beyond[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). Long Beach:IEEE, 2019.

MENG Z, KARANAM S, WU Z, et al. Re-identification with consistent attentive siamese networks[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). Long Beach:IEEE, 2019.

YAN C, PANG G, WANG L, et al. BV-person:a large-scale dataset for bird-view person re-identification[C]//2021 IEEE/CVF International Conference on Computer Vision(ICCV). Montreal:IEEE, 2021.

GU X, CHANG H, MA B, et al. Clothes-changing person re-identification with RGB modality only[C]//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). New Orleans:IEEE, 2022.

YANG M, HUANG Z, HU P, et al. Learning with twin noisy labels for visible-infrared person re-identification[C]//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). New Orleans:IEEE, 2022.

CHO Y, KIM W J, HONG S, et al. Part-based pseudo label refinement for unsupervised person re-identification[C]//2022IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). New Orleans:IEEE, 2022.

WANG H, SHEN J, LIU Y, et al. NFormer:robust person re-identification with neighbor transformer[C]//2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). New Orleans:IEEE, 2022.

WANG T, LIU H, SONG P, et al. Pose-guided feature disentangling for occluded person re-identification based on transformer[C]//Association for the Advancement of Artificial Intelligence(AAAI). Vancouver:AAAI, 2022.

YANG L, ZHANG R Y, LI L, et al. Sim AM:a simple, parameter-free attention module for convolutional neural networks[C]//International Conference on Machine Learning(ICML). Lugano:ACM, 2021.

RADENOVIC F, TOLIAS G, CHUM O. Fine-tuning CNN image retrieval with no human annotation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2019, 41(7):1655-1668.

GONG Y, WANG L, LI Y, et al. A discriminative person re-identification model with global-local attention and adaptive weighted rank list loss[J]. IEEE Access, 2020, 8:203700-203711.

HU B, XU J, WANG X. Learning generalizable deep feature using triplet-batch-center loss for person re-identification[J]. Science China Information Sciences, 2021, 64:120111.

LUO H, JIANG W, GU Y, et al. A strong baseline and batch normalization neck for deep person re-identification[J]. IEEE Transactions on Multimedia, 2020, 22(10):2597-2609.

GU H Y, FU G Y, LI J M, et al. Auto-ReID+:searching for a multi-branch ConvNet for person re-identifification[J]. Neurocomputing, 2021, 435:53-66.

浏览量

108

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

基于1D-Res&SENet的呼吸暂停检测

基于锯齿空洞残差卷积的单幅图像超分辨率重建研究