基于CNN改进模型的服装分类方法

Clothing Classification Method Based on CNN Improved Model Network

随着服装电子商务的快速发展,服装的种类日益增加,根据服装的纹理设计对其进行分类变得越来越重要。传统的图像处理方法已经很难应对越来越复杂的图片背景。为了解决这个问题,本文提出了一种改进的CNN用于服装分类方法,对原有的CNN模型进行结构的调整并在调整的结构中增加卷基层。将改进的算法应用于不同服装数据集上,对Fashion数据集的分类准确率为84.5%,对CAD数据集的分类准确率为77.8%。两个不同数据集的实验结果显示,改进的模型比现有的两个著名的CNN模型(AlexNet和VGGNet)有着更高的准确率。

With the rapid development of clothing e-commerce,the types of clothing are increasing.It is becoming more and more important to classify clothing according to their texture design.Traditional image processing methods are hard to deal with more and more complicated picture backgrounds.In order to solve this problem,this paper proposes an improved CNN for clothing classification,adjusts the structure of the original CNN model and increases the volume of the reel in the adjusted structure.The improved algorithm is applied to different garment datasets.The classification accuracy of Fashion data set is 84.5%,and the accuracy of CAD data set classification is 77.8%.The experimental results of two different datasets show that the improved model has higher accuracy than the existing two well-known CNN models(AlexNet and VGGNet).