基于深度学习的服饰属性标签识别技术

Recognition Technology of Clothing Attrib-ute Label Based on Deep Learning

摘要: 在对深度学习领域和神经网络结构深入探究及理解基础之上，针对现有方法大都基于服饰整体而忽略服饰细节要素特征的问题，本文设计并实现一种基于深度学习的服装属性标签识别的方法，采用来自阿里电商数据的网络服饰图像数据库，基于深度学习的神经网络快速高效地完成服饰图像中关键特征信息的提取，并基于提取的数据特征进行建模分析，从而达到对服饰属性的精准识别与分类。实验结果表明：该方法能有效地实现服饰属性标签的识别与分类，分类准确率有效提高，可有效解决服饰属性标签识别分类困难的问题，实现服饰图像局部属性的高效识别分类。

Abstract: On the basis of deep exploration and understanding of deep learning field and neural network structure, in view of the problem that the existing methods are mostly based on the whole dress and ignore the characteristics of dress details, in this paper, we design and implement a clothing attribute label recognition method based on deep learning, using the online clothing image data-base from Ali ecommerce data, the neural network based on deep learning can quickly and efficiently extract the key feature information in clothing image, and modeling and analysis based on extracted data features, so as to achieve accurate identification and classification of clothing at-tributes. The experimental results show that this method can effectively realize the recognition and classification of clothing attribute labels, improve the classification accuracy effectively, effectively solve the difficult problem of clothing attribute label recognition and classification, and realize the efficient recognition and classification of local attributes of clothing image.