Research On Deep Learning-based E-commerce Clothing Label Recognition

In recent years,the development of e-commerce has been rapid and has been considered an important factor in the high-quality development of the economy.In 2021,the national e-commerce transaction volume reached 42 trillion yuan,with an average annual growth rate of 20.3% from 2013 to 2021.With the continuous improvement of people’s living standards,clothing has become a fast-moving consumer product,and clothing e-commerce occupies an important position in e-commerce.The rapid product updates and the appearance of a large number of similar clothing items have caused decision anxiety among consumers,and the OEM and subcontracting models in the clothing processing industry are the main factors producing a large number of similar clothing items.Therefore,identifying OEM clothing is key to solving the problem of recommending similar clothing items.This article studies the identification of OEM clothing based on its characteristics using deep learning technology.The work of this article is as follows:Firstly,by researching the entire process from clothing processing and production to shelf sales,the characteristics of OEM clothing were summarized,namely,high similarity in clothing design,fabric composition,and size chart.Based on this,the commonality of product display pages on major e-commerce platforms such as JD and Taobao was analyzed,and their anti-crawling measures were handled.Using the Selenium+Requests framework,a spider program was developed to collect clothing product information,including clothing display images,clothing-related attribute information,and clothing details page information.Secondly,the collected clothing data was sorted and the OEM clothing recognition model was divided into three blocks of similarity calculation based on the characteristics of OEM clothing.The similarity of the clothing design was measured by the similarity of the clothing product image.The VGG16 model was used to extract the features of the product image,and the cosine distance was used to measure the similarity of the image feature vectors.The size chart was obtained by training an object detection model to extract it from the product detail page,and then using table recognition technology to convert the size chart from an image to text stored in an Excel table.The similarity of the size chart was measured by normalizing it and calculating the sum of the differences in corresponding attributes.The similarity of the fabric information was calculated by rule matching,with a similarity value of 0 for consistent fabric information and 1 for inconsistent information.The OEM clothing recognition model was given by combining the three similarity values.Finally,relying on the OEM recognition model,a clothing OEM recognition system was built.This system not only has the functions of traditional e-commerce platforms but also has the function of identifying OEM clothing,improving the user experience.