Sample set construction method and device, computer equipment and storage medium

Abstract

The invention relates to a sample set construction method and device, computer equipment and a storage medium. The method relates to sample generation for training a model. The sample set constructionmethod comprises the steps of obtaining a certificate template graph which is generated according to a certificate graph and does not comprise certificate information; generating a plurality of groups of virtual certificate information according to the styles of various certificate information in the certificate graph; writing the virtual certificate information into a certificate template graphaccording to the position of each type of certificate information in the certificate graph to generate an electronic certificate graph; carrying out image collection on the entity certificate corresponding to the electronic certificate picture to obtain a certificate collection picture; and constructing a picture sample set according to the electronic certificate image and the certificate collection image, wherein the picture sample set is used for training a character recognition model. By adopting the method, the balance of the samples for training the character recognition model can be improved.

Description

technical field [0001] The present application relates to the field of computer technology, in particular to a sample set construction method, device, computer equipment and storage medium. Background technique [0002] In the technical field of automatic identification of document information, a large number of document pictures are required to train the character recognition model, which can improve the accuracy of the character recognition model in identifying document information. However, the number of pictures required for training character recognition models is very large, and it is usually impossible to obtain a large number of real document pictures. [0003] At present, most of the ID images used in training character recognition models are obtained by batch-generating electronic ID images from templates. Although the electronic ID images obtained in this way are relatively clear, their randomness is not strong, which will cause ID images to exist as sample images...

AI Technical Summary

Problems solved by technology

However, the number of pictures required for training character recognition models is very large, and it is usually impossible to obtain a large number of real ID pictures

[0003] At present, most of the ID images used in training character recognition models are obtained by batch-generating electronic ID images from templates. Although the electronic ID images obtained in this way are relatively clear, their randomness is not strong, which will cause ID images to exist as sample images. The problem of uneven distribution

When directly using a large number of unbalanced certificate pictures to train the character recognition model, the model parameters of the obtained character recognition model will not be accurate enough. When the trained character recognition model is used to recognize the certificate information, the recognition results obtained will not be accurate too accurate

Images