Research On Clothing Classification And Recommendation Methods Based On Image Content

With the increasing demand of clothes,adapting computer vision method to the studies of clothing field has become a heated point to study.It's difficult to recommend clothing personalized since the high variability and complexity of clothing,the accuracy of existing recommendation results need improve.As for the problem of clothing color not in keeping with people's color of skin,eyes,hair and clothing styles,a new clothing recommendation method based on Image content is brought about.As for the high variability and complexity of clothing,this paper shows a new method of clothing classification on the bases of feature extraction and distance measurement.First,establish clothing images Data Set.Secondly,mark styles and properties according to the pitchers by Mechanical Turk.Third,estimate all the pose and extract the feature shown in the pitchers.Finally,classify multiple clothing attribute label according to feature extraction and distance measurement,implement clothing classification bases of different clothing styles.As for the problem of clothing color not in keeping with people's color features and low accuracy of existing recommendation results,combining Color Theory of Four Seasons and computer vision.This paper also brought a new clothing recommendation method based on Image content.First,four seasons color judgment model is brought out and classify people's color features picked up from facial images.And then build a optimization model.Optimize the styles of the clothing and gain the result of pre recommendation according to the result of four seasons color judgment model and people's preference.Finally,improve the result and find the final recommendation result according to users rating and feedback.Experiments show that the new method of clothing classification on the bases of feature extraction and distance measurement brought about in this paper can classify different styles of clothes.The method of clothing classification on the bases of feature extraction and distance measurement can adapt clothes colors to people's color features and it's recommendation results have higher accuracy in application.