Identification Of Luxury Based On The Similarity Measurement Of Siamese Networks

In recent years,domestic consumption of luxury goods has repeatedly hit new highs.With the continuous increase in the volume of luxury goods trade,the authenticity of luxury goods has gradually become a stumbling block that plagues the development of the industry.Traditional solutions that rely on experts for manual identification have also shown their own bottlenecks as the magnitude of luxury goods authenticity identification increases.Manual identification is not only time-consuming,labor-intensive,and costly.In response to this problem,this paper proposes a model that utilizes deep learning technology to tackle the task of luxury goods authenticity identification based on the visual characteristics of the luxury goods,and develops an online system for mobile-end luxury goods authenticity identification tasks with the model —— Real identification of We Chat applet.The problem studied in this paper is how to build a method for efficiently identifying the authenticity of luxury goods on the mobile phones,so that users can quickly and easily identify the authenticity of luxury goods through mobile phones and cameras.The difficulty of the mobile luxury authenticity identification task lies in the complexity of luxury categories,diverse appearances,and difficulty in dataset construction.In the application process,images of luxury goods are taken very casually,and the quality of these images is difficult to unify standards,which results in that the visual characteristics of luxury goods are weak and easy to be disturbed and destroyed.In response to the above problems,this paper firstly constructs a dataset called LGID(Luxury Goods Identification Dataset)that can be used for luxury goods authenticity identification tasks,which provides a benchmark for the research of deep learning in the field of luxury goods.Then this paper investigates and discusses the feasibility of luxury authenticity identification,and puts forward the core hypothesis of luxury authenticity identification based on visual abnormalities.Subsequently,this paper proposes a framework with Siamese Network as the backbone and combining with Spatial Pyramid Pooling to identify the authenticity of luxury goods.This method adopts the twin network contrast learning idea,showing good robustness for identifying luxury goods under micro-abnormalities,and finally achieves 95.4% accuracy in the dataset of 21459 pictures of 52 samples of LV package.In the process of identification application,in view of the problem that the luxury goods images are taken too casually,or the images are compressed or damaged during the image transmission process,the authenticity information of the images is always lost.This paper also proposes a filter method that can be utilized to determine whether the image is blurred,whether the image resolution is too low and whether the quality of the luxury goods image is satisfied for luxury goods authenticity identification.The method is composed by an image classification module,a special pattern detection module,and a resolution definition filtering module,which provides high-quality samples for the luxury goods authenticity identification model.Finally,this paper combines the luxury identification method and filter method proposed above to construct an online system We Chat applet Real identification that can be used for luxury authenticity identification on mobile phones.The system allows users to take the images of the luxury goods according to the samples on the We Chat applet,upload these images to the cloud server for identifying the authenticity of the luxury goods.The system finally feedbacks the authenticity of the sample to the users.