The Design And Implementation Of Clothing Image Retrieval System Based On HOG And SVM

According to the ‘2012Annual Report of Chinese clothing e-commerceoperation’, the scale of China’s online clothing shopping is increasing year by yearsince2009. The growth rate of online clothing shopping in2010is100.8percent,which is the most rapid. In2012, the trading volume throughout China online clothingshopping market had reached305billion Yuan. It is expected that the transaction sizeof online clothing shopping market in China will reach407.6billion Yuan. The reportshows that clothing e-commerce has a huge potential for space and unparalleleddevelopment in China.The development of e-commerce clothing is associated with the development ofimage retrieval technology. However, the current majority of online clothing sites,areusing the text-based image retrieval technology. This technology is only use thekeywords like ‘red’’dress’ to retrieval cloths, which can not offer the clothing styleappearance properties to satisfy the users.In order to allow users to intuitively user visual image information to retrieveclothing products, Fashion Retrieval System come into being. Fashion retrievalsystem dedicated to improve the user experience of online clothing shopping, provideusers with a new online retrieve platform in the way of ‘from image to image’The main contribution of this paper is the design and implementation of anonline clothing image retrieval platform: Fashion retrieval system. This system usescontent-based image retrieval methods to provide the search function with imageinput to image output. Users can upload images of fashion clothing, and then search for similar clothing images. Users can compare the differences between two similarproducts and find the most suitable one.This paper also analyzes the work and research of the current image searchtechnology, and find that using HOG feature+SVM trainer framework has the bestresults in clothing images search by experiment.This paper use the thought of engineering of software to make a deep andcomprehensive analysis to the Fashion Retrieval System. It use the current-popularWeb application framework that implements a clothing image searching platform, andthe overall code lines reach15000. The main function of the Fashion RetrievalSystem has been tested completely and is runnable,which now has online in ourLaboratory.