A personalized clothing recommendation method based on dual-modal content analysis of clothing images and label text

Abstract

The invention discloses a personalized clothing recommendation method based on dual-modal content analysis of clothing images and label text, which includes the following steps: S1: By analyzing the detailed attributes in the clothing images of the shopping website, establishing a detailed location of clothing Clothing product model and user preference model described in the form of images and features; S2: By analyzing the clothing label text of the shopping website, establish a clothing product model and user preference model described in text form; S3: Establish the clothing product model and user preference model established in step S1 The clothing product model based on clothing images is combined with the user preference model established in step S2 to generate recommendation results. The present invention combines clothing images and text information, and makes personalized clothing recommendations for users based on the fusion of the two aforementioned models.

Description

technical field [0001] The invention relates to a method for recommending personalized clothing based on bimodal content analysis of clothing images and label texts, which belongs to clothing recommendation methods. Background technique [0002] With the development of Internet technology, more and more e-commerce websites have appeared, selling various types of goods, including cosmetics, clothing, Haitao, daily necessities, electronic products, etc. Online shopping has almost become a must for every family. less part. The reason why shopping websites are so popular is that this massive "commodity library" includes commodities from all over the country and even the world. However, while users enjoy this convenience, they also face such a problem: when they visit e-commerce websites, it is difficult to directly find their target products at the beginning, and they will passively browse a lot of product information that they do not care about. It takes a lot of time to find...

AI Technical Summary

Problems solved by technology

However, while users enjoy this convenience, they also face such a problem: when they visit e-commerce websites, it is difficult to directly find their target products at the beginning, and they will passively browse a lot of product information that they do not care about. It takes a lot of time to find information about the products you need

However, clothing products have some particularities: first, users generally do not buy the same clothing multiple times; second, each user has their own dressing preferences, and most of them don’t like to follow the crowd; third, text labels often cannot accurately describe details of clothing

Due to the user's "non-conformity" psychology, the collaborative filtering recommendation method is not suitable for clothing recommendation, and there are a large number of new clothing on the business website every day or even every moment. For a new clothing that has just been released, no user has ever purchased it. Then there is no way to recommend it to users who may be interested in it through collaborative filtering

In addition, due to the "no repetition" psychology of users when purchasing clothing, traditional content-based recommendation methods are not suitable for clothing recommendation

Images