摘要
随着智能化时代的来临,社会智能监控设备全面覆盖,行人重识别成为极具挑战的研究课题。行人重识别的核心在于设计深度学习网络对图像提取强判别力的特征,本文针对局部特征的重要性提出一种局部特征融合的行人重识别模型。将一个卷积块接出2条独立支路,分别接入局部特征提取模块和批特征擦除模块,最后和全局特征提取支路进行特征融合获得高细粒度的特性,并采用联合损失函数训练网络。在Marketl501数据集上验证所提方法具有有效性,mAP达到84.12%,Rank-1达到95.06%。
With the advent of the intelligent era and the full coverage of social intelligent monitoring equipment,pedestrian re-recognition has become a very challenging research topic.The core of pedestrian re-recognition is to design a deep learning network to extract strong discriminant features from the image.Aiming at the importance of local features,this paper proposes a pedestrian re-recognition model based on local feature fusion.A convolution block is connected to two independent branches,which are respectively connected to the local feature extraction module and the batch feature erasure module.Finally,the features are fused with the global feature extraction branch to obtain the characteristics of high fine granularity,and the joint loss function is used to train the network.The effectiveness of the proposed method is verified on marketl501 dataset,with mAP of 84.12%and rank-1 of 95.06%.
作者
涂园园
贺松
姚绍华
TU Yuanyuan;HE Song;YAO Shaohua(School of Big Data and Information Engineering,Guizhou University,Guiyang 550025,China)
出处
《智能计算机与应用》
2021年第12期122-125,132,共5页
Intelligent Computer and Applications
关键词
局部特征
批特征擦除
深度学习
损失函数
local features
batch feature erasure
deep learning
loss function
