基于卷积神经网络的服装种类识别

Classification of Clothing Type Based on Convolutional Neural Network

随着网络购物和电子商务的普及,越来越多的电商平台推出了移动端的应用,这其中不少都集成图片搜索的功能,根据输入的图片判断商品的种类进行搜索。基于卷积神经网络的服装种类识别正是为了解决其中对服装的分类而提出的方法,具体包括数据预处理、卷积运算和Softmax回归模型训练三个步骤。在来自真实购物平台的服装图片组成的实验数据集上,能够保持较高的识别准确率,在服装识别问题上具有很大的优势。

With the prevalence of online shopping and Electronic Commerce, increasing number of e-commerce companies have released their applications embedded on mobile devices, some of which are integrated with image searching function. Users can search kinds of goods with input image. Classification of clothing type based on convolutional neural network rightly resolves the classification problems on clothing types. This method contains three steps： pre-processing, convolution and Softmax regression. It shows a good classification accu-racy on the dataset collecting from reality shopping platform, which makes it have great strength on clothing type recognition.