A method for identifying cigarette box gauge based on deep learning technology

Abstract

The invention discloses a cigarette box gauge identification method based on deep learning technology, which comprises the following steps: step S1, collecting and screening pictures to form a picturelibrary; collecting and arranging the application scenes of cigarette box pictures, and collecting different kinds of cigarette box pictures in different states to form a picture library; step S2, building convolutional neural network; step S3, training that model; step S4, pre-judging the model, loading the trained model, inputting the cigarette box picture to be recognized, and judging whetherthe model is valid or not; step S5, deploying that model; step S6, providing the service and deploying it to the cloud server. The method has the advantages of high identification accuracy, easy popularization, wide application range and the like.

Description

[0001] Full process in the technical field [0002] The present invention mainly relates to the technical field of management of cigarette production, sales and other links, in particular to a method for identifying cigarette pack specifications based on deep learning technology. Background technique [0003] At present, in the production and sales of cigarettes, cigarette boxes are generally used as identification features of cigarette categories and specifications. Therefore, the identification technology of cigarette packs directly affects the management accuracy and efficiency of cigarette production and sales links. The identification of traditional cigarette packs generally adopts the following three methods: [0004] 1. Manual identification: The disadvantage of manual identification is that it is easy to confuse cigarette product regulations information for cigarette retailers and consumers when the appearance of cigarette packs is not much different. [0005] 2. Bar...

AI Technical Summary

Problems solved by technology

[0004] 1. Manual identification: the disadvantage of manual identification is that it is easy to confuse cigarette product regulations information for cigarette retailers and consumers when the appearance of cigarette packs is not much different

[0005] 2. Barcode recognition: The disadvantage of barcode recognition is that it can only partially display the name of the cigarette brand, but cannot display the complete basic information and brand culture of the cigarette brand, and the query method is single. If part of the barcode is missing, it cannot be recognized

[0006] 3. Two-dimensional code recognition: The disadvantage of two-dimensional code recognition is that only a small number of cigarette boxes on the market have two-dimensional code design, and it is mainly used for the promotion activities of various China Tobacco companies, which is not universal

[0007] In the field of cigarette pack image recognition, deep learning has not been introduced for recognition. The existing picture library has very few pictures of cigarette brands, which requires a lot of work to collect. There is no algorithm specifically for cigarette pack recognition, and the algorithm cannot be applied mechanically. , it is necessary to develop a corresponding algorithm to train the collected pictures

Images