Design And Implementation Of Clothing Image Auto-tagging System Based On Deep Learning

With the rapid development of E-commerce, more and more traditional offline clothing sellers begin to sell their product on the Internet. When entering information of the clothes, the sellers need to tag the clothes manually. This manual tagging way is okay in the past, but now, with the explosive growth of clothes on the Internet, it waste lots of time to tag so many products. In the meantime, with the developing of machine learning, researchers in the field of deep learning has made great breakthrough. In many dataset of image classification, the method using deep learning is the state of art. Many company start to focus on the research of deep learning, including Google, Fackbook, Amazon and Baidu etc.Because the huge success of deep learning in image classification, this paper try to develop a clothing image auto-tagging system based on deep learning. This paper design and implement a multi-label classifier and implement an image auto-tagging web application system based on the classifier. The system can deal with image that have arbitrary number of tags. In the dataset that crawling from the jingdong.com, the system achieve a 79.256% accuracy. Our system can be embed into clothing E-commerce website and help the seller to auto generate image tags when they upload an image. Even though, the system is designed based on clothing image, it can be extend easily to other kinds of product.