Clothing Recommendation Svstem Based On MASK Detection

With the rapid development of e-commerce,the number of clothing images is increasing on shopping platforms,and clothing styles are diverse.A single text search of clothing images has long been unable to meet user needs.For clothing multi-tag image indexing tasks,the processing and analysis of traditional techniques rely on manual labeling of clothing image information.With the explosion of data,its shortcomings are increasingly apparent,which has also promoted the development of images in the field of retrieval and recommendation.The application research of clothing recommendation system based on Mask detection is a systematic research on deep learning technology,image classification and retrieval technology.In this paper,an improved deep multi-similar hash model is proposed.On the basis of the DMSH model,weighted loss,bilateral zero loss thresholds are added,and a strategy that conforms to the model is proposed.According to the importance of tags,weights are introduced in the model.The system's multi-process retrieval recommendation algorithm can achieve multi-user retrieval in the millions of data sets in milliseconds,and obtain satisfactory Top-n recommended images.In order to eliminate as much as possible the influence of irrelevant information in the clothing image on the recommendation system,the image segmentation uses Mask R-CNN network.In order to train the image segmentation model and the DMSH model well,two data sets were produced,and the data sets were described in detail.In this paper,the structure of the system is recorded in detail,and the accuracy of the clothing category Top-5 has reached 86.4%.