Image generation model and card number identification model training method and device

Abstract

The invention provides an image generation model and a card number identification model training method and device, and the method comprises the steps of carrying out the processing of a sample three-dimensional card image comprising a first sample card number in a three-dimensional form, and obtaining a sample background image and a sample plane card number image comprising a first sample card number in a plane form; taking the sample background image and the sample plane card number image as a first input image of an image generation model; and determining first target loss according to theoutput image and the sample three-dimensional card image, and adjusting a first parameter corresponding to the image generation model according to the first target loss until the output image of the image generation model is matched with the sample three-dimensional card image to obtain an image generation model comprising the trained first parameter. According to the invention, the three-dimensional card image with the card number in the three-dimensional form can be generated, the requirement for the training sample is simplified, the three-dimensional card image can serve as the training sample to train the card number recognition model, and the recognition accuracy of the card number recognition model is improved.

Description

technical field [0001] The present application relates to the technical field of machine learning, in particular, to a training method and device for an image generation model and a card number recognition model. Background technique [0002] In daily life, users usually hold multiple cards of different types, such as bank cards, membership cards, and social security cards. Users can usually carry out various activities through the card, such as transaction payment, transaction points, information inquiry and so on. [0003] With the rapid development of Internet technology, user activities are currently carried out online. In this way, users are required to bind matching cards on the corresponding activity platform in advance. Make transaction payments. At present, the user manually enters the card number to bind the card on the activity platform, which is inefficient. [0004] Based on this, the card number recognition model is trained to automatically recognize the car...

AI Technical Summary

Problems solved by technology

At present, the user manually enters the card number to bind the card on the event platform, which is inefficient

[0004] Based on this, the training card number recognition model is used to automatically recognize the card number through the card number recognition model. However, the recognition accuracy of the currently trained card number recognition model is poor and cannot meet the actual needs.

Images