Research Of Key Point Based Style Recognition And Intelligent Fashion Clothing Recommendation System

With the rapid growth of development in the Internet, more and more customers are willing to invest their fashion choice across the Internet. But it is still very hard for customers to find their personalized fashion choice in mass data efficiently. To fulfill this gap, some recommendation systems have been proposed, but for most of these systems extract the features of clothing by hand, they can only run in local database, and is hard to adapt the mass fashion data analysis in the Internet. Since fashion information is often presented by visual method, if a system can properly realize fashion multimedia mining in the Internet, the consumers can get the personalized and diversity choice in the Internet at very first time, but it is still a challenge.With help of the proposed visual methods, we realize an intelligent personalized fashion recommender based on clothing multi-feature to connect online real-time multimedia mining with recommendation. The advantages of the proposed system are listed as follows:(i) Oriented to the Internet proposed system associates clients’ personalized and diversified demand with the current fashion trend based on visual methods efficiently and accurately.(ii) Key points based clothing style recognition module gives a novel way to define the clothing style.(iii) Evolutionary hierachical fashion multimedia mining creates a hierachical structure to filter the key components of fashion multimedia information from the Internet, and a refined contour extraction of fashion model method is proposed to adapt the dynamic and complex environment.(iv) A multi-dimensional clothing feature description module is proposed in the system, which considers main color tone, skin color and style.Our research helps customers to find their personal demand in fashion trend of the Internet, which connects the customers, professional designers, and clothing manufacturers together. There has a great market development potential in the proposed system.