Research And Implementation Of Clothing Image Retrieval Algorithm Based On Deep Learning

In recent years,there has been a great development in the field of e-commerce.The online transaction volume of the clothing industry has continued to expand as a proportion of the total online shopping transaction.The types and quantity of clothing provided by e-commerce platforms are also growing rapidly,which has led to ordinary image retrieval engines.It is difficult to accurately retrieve the clothing products that users need.At present,clothing retrieval mainly uses two methods:text information and image content.The text-based clothing retrieval method has disadvantages: clothing images are rich in information,it is difficult to describe the content of the images in detail with only a few texts,and different individuals have different text descriptions of clothing images.Traditional content-based clothing retrieval has high requirements for image features.Researchers need to have rich experience to extract good image features.In view of the wide application of deep learning in recent years,it has practical significance to explore clothing image retrieval methods based on deep learning.On the basis of summarizing the current research status of clothing image retrieval,this paper designs a DMC-Mobile network based on a lightweight neural network in view of many difficulties such as deformation,illumination,and wrinkles in clothing images.The main research work of this paper includes the following points:(1)A model structure based on a lightweight neural network is designed.The overall retrieval framework combines the characteristics of clothing images and dynamic alignment algorithms.At present,the number of clothing images is huge.The deep learning model can be used to extract clothing image features simply and efficiently.In addition,during model training The triple loss is added,and the K-Means clustering algorithm is added in the image retrieval stage,which greatly accelerates the retrieval speed,A comparative experiment of clothing image retrieval of classic network models Google Net,VGG,Mobile Net and DMC-Mobile is carried out.The experimental results show that DMC-Mobile has higher accuracy and retrieval speed than common convolutional neural network models.(2)Designed and implemented a clothing retrieval system software with front and back ends,which has a certain practicability.