ID 原文 译文
38886 为此，本文尝试将EfficientNet系列网络引入到行人重识别领域，替代比较流行的ResNet50主干网络，提供了一个全新的骨干网基线。 In this paper, we try to introduce the EfficientNet series network into the area of Person Re-Identification, replace the more popular ResNet50 trunk network, and provide an entirely new baseline network.
38887 本文重点根据EfficientNet系列网络给出一种二分支行人重识别网络构造。 Then, we propose to design a two-branch EfficientNet for Person Re-Identification.
38888 相比于ResNet50，基于EfficientNet构造的二分支行人重识别网络具有网络参数规模小、性能提升明显的特点。 Compared with various ResNet-50-based solutions, the proposed EfficientNet has a significantly small size on network, but often achieves better performance.
38889 实验结果表明:所构造的网络在行人重识别流行数据集上均有良好的表现。 Experimental results show that the proposed network performs very well on the popular Person Re-Identification data set.
38890 为了提高指静脉描述子的鲁棒性同时降低网络参数量，通过修改VGGFace-Net并引入局部聚合描述子向量(Vector of Locally Aggregated Descriptors，VLAD)得到参数量仅0.3M的指静脉图像描述子提取网络，VLAD编码过程实现了局部描述子的聚类和重组，使描述子对手指姿势变化更加鲁棒; In order to increase the robustness of finger vein descriptors and reduce the number of network parameters, we proposed a way of modifying the VGGFace-Net and using the Vector of Locally Aggregated Descriptors(VLAD). The number of parameters of our obtained network is only 0.3 M. The VLAD can cluster and rearrange the local descriptors, which makes our descriptors more robust against the changes of finger posture.
38891 由于公开的指静脉训练数据库规模通常不够大，提出基于三元组和难分负样本挖掘策略进行网络的训练，并针对三元组损失没有约束样本对距离的类内方差的问题，提出一种样本对中心约束损失函数，通过将正负样本对视为两个类别，进一步促使其靠近各自的类中心，从而增大类内紧凑程度。 Since the size of public finger vein databases is small, we trained the network by the triplet with the hard negative sample mining strategy. However, the Triplet Loss does not constrain the intra-class variance of the sample pair distance, we further proposed an improved loss called Pair-center-constrained loss.
38892 在三个公开数据库FV-USM，SDUMLA，MMCBNU上的指静脉验证结果表明，所提取的描述子在基于欧氏距离进行匹配的情况下，指静脉验证的结果均优于现有方法，且在图像发生随机平移时具有更好的鲁棒性。 By treating positive and negative sample pairs as two categories, we can drive them even further to their class centers, which can increase the intra-class compactness. Experimental results on three public databases FV-USM, SDUMLA, and MMCBNU show that the proposed method is better than two state-of-the art methods in terms of accuracy. Meanwhile, our finger vein descriptors have better robustness against random translation.
38893 三维卷积神经网络比二维卷积神经网络具有更优越的时空特征提取能力，但运算量却显著增加。 3D convolutional neural network has superior ability in spatio-temporal feature extraction than 2D convolutional neural network, but the calculation intensity is significantly increased.
38894 针对如何有效减少模型参数量、解决准确率随着计算复杂度降低而降低的问题，提出基于端到端的通道可分离卷积神经网络。 To solve the problem of declined precision caused by reducing the computing complexity, the efficient compression of the model parameters is the key. Hence, an end-to-end channel separable convolutional neural network is proposed.
38895 通过分离通道交互作用和时空交互作用来分解三维卷积，其中分别利用3×3×3 Depthwise卷积和1×1×1常规卷积进行分离通道交互作用和时空交互作用。 3D convolution is decomposed by separating channel interaction and spatio-temporal interaction, in which 3×3×3D epthwise convolution and 1×1×1 conventional convolution are respectively used to separate channel interaction and spatio-temporal interaction.