基于多级注意力跳跃连接网络的行人属性识别

doi:10.19678/j.issn.1000-3428.0057107

摘要/Abstract

摘要： 行人属性识别易受视角、尺度和光照等非理想自然条件变化的影响，且某些细粒度属性识别难度较大。为此，提出一种多级注意力跳跃连接网络MLASC-Net。在网络中间层，利用敏感注意力模块在通道及空间维度上对原特征向量进行筛选加权，设计多级跳跃连接结构来融合所提取的显著性特征。在网络顶层，改进多尺度金字塔池化以融合局部特征和全局特征。在网络输出层，结合验证损失算法自适应更新损失层，加速模型的收敛并提高精度。在PETA和RAP数据集上的实验结果表明，MLASC-Net的识别准确率相较原基准网络分别提高约4.62和6.54个百分点，其在识别效果和模型收敛速度上有明显优势，同时在非理想自然条件下具有良好的泛化能力，可有效提高网络对细粒度属性的鲁棒性。

关键词: 行人属性识别, 多级跳跃连接网络, 敏感注意力, 多尺度金字塔, 残差网络

Abstract: Pedestrian attribute recognition is susceptible to changes in non-ideal natural conditions,including perspective,scale and lighting,which increases the difficulty in fine-grained attribute recognition.To address the problem,this paper proposes a multi-level attention skip connection network,MLASC-Net.In the middle layer of the network,the sensitive attention module is used to filter and weight the original feature vectors in the channel and spatial dimensions,and a multi-level skip connection structure is designed to fuse the extracted salient features.In the top layer of the network,multi-scale pyramid pooling is improved to integrate local features and global features.In the output layer of the network,the verification of the loss trend algorithm is used to update the loss layer adaptively,accelerate the convergence of the model and increase accuracy.Experimental results on PETA and RAP datasets show that compared with the original benchmark network,MLASC-Net improves the recognition accuracy by about 4.62 and 6.54 percentage points,respectively.It has obvious advantages in recognition effect and model convergence speed,and has good generalization ability under non-ideal natural conditions,which can effectively improve the robustness of the network to fine-grained attributes.

Key words: pedestrian attribute recognition, multi-level skip connection network, sensitive attention, multi-scale pyramid, Residual Network (ResNet)

中图分类号:

TP391.41

王林, 李聪会. 基于多级注意力跳跃连接网络的行人属性识别[J]. 计算机工程, 2021, 47(2): 314-320.

WANG Lin, LI Conghui. Pedestrian Attribute Recognition Based on Multi-Level Attention Skip Connection Network[J]. Computer Engineering, 2021, 47(2): 314-320.

http://www.ecice06.com/CN/Y2021/V47/I2/314

图/表 10

20210225093430

20210225093432

20210225093435

20210225093439

20210225093442

20210225093445

20210225093448

20210225093451

20210225093454

20210225093457

参考文献 22

[1]	SUDOWE P,SPITZER H,LEIBE B.Person attribute recognition with a jointly-trained holistic CNN model[C]//Proceedings of 2015 IEEE International Conference on Computer Vision.Washington D.C.,USA:IEEE Press,2015:87-95.
[2]	ZHANG N,PALURI M,RANZATO M,et al.PANDA:pose aligned networks for deep attribute modeling[C]//Proceedings of 2014 IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C.,USA:IEEE Press,2014:1637-1644.
[3]	FABBRI M,CALDERARA S,CUCCHIARA R.Genera-tive adversarial models for people attribute recogni-tion in surveillance[C]//Proceedings of 2017 IEEE International Conference on Advanced Video and Signal Based Surveillance.Washington D.C.,USA:IEEE Press,2017:1-6.
[4]	SUDOWE P,LEIBE B.PatchIt:self-supervised network weight initialization for fine-grained recognition[C]//Pro-ceedings of British Machine Vision Conference.New York,USA:ACM Press,2016:24-30.
[5]	LI Y N,HUANG C,LOY C C,et al.Human attribute recognition by deep hierarchical contexts[M].Berlin,Germany:Springer,2016.
[6]	ZHU Jianqing,LIAO Shengcai,YI Dong,et al.Multi-label CNN based pedestrian attribute learning for soft biometrics[C]//Proceedings of 2015 International Conference on Biometrics.Washington D.C.,USA:IEEE Press,2015:535-540.
[7]	LI Dangwei,ZHANG Zhang,CHEN Xiaotang,et al.A richly annotated dataset for pedestrian attribute recognition[EB/OL].[2019-12-10].https://arxiv.org/abs/1603.07054.
[8]	WANG Feng,LIU Qingshan.Fine pedestrians recognition based on double convolutional neural network[J].China Sciencepaper,2017(14):17-21.(in Chinese)王枫,刘青山.基于双卷积神经网络的行人精细化识别[J].中国科技论文,2017(14):17-21.
[9]	LIU Xihui,ZHAO Haiyu,TIAN Maoqing,et al.HydraPlus-Net:attentive deep features for pedestrian analysis[C]//Proceedings of 2017 IEEE International Conference on Computer Vision.Washington D.C.,USA:IEEE Press,2017:350-359.
[10]	SARAFIANOS N,XU X,KAKADIARIS I A.Deep imbalanced attribute classification using visual attention aggregation[M].Berlin,Germany:Springer,2018.
[11]	SAQUIB M S,SCHUMANN A,WANG Y,et al.Deep view-sensitive pedestrian attribute inference in an end-to-end model[C]//Proceedings of British Machine Vision Conference.Washington D.C.,USA:IEEE Press,2017:159-168.
[12]	HE Kaiming,ZHANG Xiangyu,REN Shaoqing,et al.Deep residual learning for image recognition[C]//Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C.,USA:IEEE Press,2016:770-778.
[13]	DENG Y B,LUO P,LOY C C,et al.Pedestrian attribute recognition at far distance[C]//Proceedings of the 22nd ACM International Conference on Multimedia.New York,USA:ACM Press,2014:789-792.
[14]	LI Dangwei,ZHANG Zhang,CHEN Xiaotang,et al.A richly annotated pedestrian dataset for person retrieval in real surveillance scenarios[J].IEEE Transactions on Image Processing,2019,28(4):1575-1590.
[15]	HUANG Zhanchao,WANG Jianlin.DC-SPP-YOLO:dense connection and spatial pyramid pooling based YOLO for object detection[J].Information Sciences,2020,522:241-258.
[16]	HU J,SHEN L,SUN G.Squeeze-and-excitation networks[C]//Proceedings of 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C.,USA:IEEE Press,2018:7132-7141.
[17]	WOO S,PARK J,LEE J Y,et al.CBAM:convolutional block attention module[M].Berlin,Germany:Springer,2018.
[18]	HE Keke,WANG Zhanxiong,FU Yanwei,et al.Adaptively weighted multi-task deep network for person attribute classification[C]//Proceedings of the 25th ACM International Conference on Multimedia.New York,USA:ACM Press,2017:1636-1644.
[19]	LI Dangwei,CHEN Xiaotang,HUANG Kaiqi.Multi-attribute learning for pedestrian attribute recognition in surveillance scenarios[C]//Proceedings of 2015 IAPR Asian Conference on Pattern Recognition.Washington D.C.,USA:IEEE Press,2015:111-115.
[20]	LI Dangwei,YU Kai,LENG Biao,et al.Weakly-supervised learning of mid-level features for pedestrian attribute recognition and localization[C]//Proceedings of British Machine Vision Conference.Washington D.C.,USA:IEEE Press,2017:15-26.
[21]	WANG Jingya,ZHU Xiatian,GONG Shaogang,et al.Attribute recognition by joint recurrent learning of context and correlation[C]//Proceedings of 2017 IEEE Inter-national Conference on Computer Vision.Washington D.C.,USA:IEEE Press,2017:531-540.
[22]	ZHAO Xin,SANG Liufang,DING Guiguang,et al.Grouping attribute recognition for pedestrian with joint recurrent learning[EB/OL].[2019-12-10].https://www.ijcai.org/Proceedings/2018/0441.pdf.

选择文件类型/文献管理软件名称

选择包含的内容