Picture screening method and device and electronic equipment

Abstract

The invention discloses a picture screening method and device, and electronic equipment. The picture screening method comprises the steps of obtaining to-be-screened pictures; inputting the to-be-screened pictures into a text picture recognition model to perform recognition processing on a picture area containing text information to obtain a recognition result; when a picture area containing textinformation exists in the recognition result, selecting a to-be-screened picture as a target picture, wherein the character and picture recognition model comprises a shallow neural network model determined by performing machine learning based on feature data output on the basis of a target deep neural network model, input picture corresponding to the output, and label information corresponding tothe input picture. According to the method, whether the picture area in the picture contains the text information or not is recognized based on the shallow neural network model with the deep neural network model prediction capability, namely the text picture recognition model, so that the picture screening speed is greatly increased, and the screening stability and screening precision are guaranteed.

Description

technical field [0001] The invention relates to the field of computer technology, in particular to a picture screening method, device and electronic equipment. Background technique [0002] The text in the picture often carries important information, but often only a small part of the pictures contain text information in a large number of pictures that need to be processed. In order to reduce the burden on the data processing system, the pictures can be filtered to filter out pictures without text information , and only process pictures containing text information. [0003] In related technologies, it is generally based on hand-designed Unicom region features to identify whether there is a Unicom region containing text information in a picture. Since the manually designed Unicom region features are not fully guaranteed in terms of classification and robustness, so that The picture screening method in the related art has poor stability and screening accuracy; in addition, be...

AI Technical Summary

Problems solved by technology

[0003] In related technologies, it is generally based on hand-designed Unicom regional features to identify whether there is a Unicom area containing text information in the picture. and robustness are not fully guaranteed, so that the stability and screening accuracy of the picture screening method in the related art are relatively poor; in addition, because the algorithm for determining the characteristics of the Unicom area in the picture and the Unicom area is usually more complicated , which makes it difficult for the picture screening method in the related art to achieve fast response

Images