结合嵌入模块的细粒度图像分类方法

A Fine-Grained Image Classification Method Combined With an Embedded Module

为提高细粒度图像分类任务的准确性,提出结合嵌入模块并基于双线性网络(Bilinear CNN)的改进方法。首先,选取VGG-16前5个卷积块,以及ResNet-50前7个卷积块分别作为两路特征提取模块。并在两个分支的特征提取器中分别嵌入自适应权重特征模块(AWFF)和特征重标定模块(SENet Block)。借助AWFF模块实现深层与浅层特征的融合,通过SENet Block实现特征通道的重标定功能,将重点集中在主要特征上。经验证,该方法能提高模型分类的准确度可达96.362%,有效提高工作效率。

In order to improve the accuracy of fine-grained image classification,an improved method combining embedded module and Bilinear Net⁃work(Bilinear CNN)is proposed.First,the first 5 convolution blocks of VGG-16 and the first 7 convolution blocks of ResNet-50 are select⁃ed as two feature extraction modules respectively.Adaptive weight feature module(AWFF)and feature re-calibration module(SENet block)are respectively embedded in the feature extractors of the two branches.The AWFF module is used to realize the fusion of deep and shallow features,and SENet block is used to realize the re-calibration function of feature channels,focusing on the main features.It is proved that this method can improve the accuracy of model classification up to 96.362%and improve the work efficiency effectively.