Picture screening method, system and equipment, and storage medium

Abstract

The invention provides a picture screening method, system and equipment, and a storage medium. The method comprises the steps of obtaining to-be-tested pictures; calculating a first quality score of the to-be-tested picture by using a target single-task network model, wherein the target single-task network model is obtained by training based on a target multi-task network model, wherein the targetmulti-task network model is obtained by training a first sample image and a label value of each first sample image under at least one preset attribute class; and screening the to-be-tested pictures according to the first quality score. According to the invention, the size of the model is reduced, the model training efficiency and the operation speed are improved, and the picture screening efficiency is improved.

Description

technical field [0001] The present invention relates to the field of computer technology, in particular to a picture screening method, system, equipment and storage medium. Background technique [0002] In the face recognition system, for example, when comparing faces, it may be necessary to collect a large number of face images of the target user, and then compare the collected face images with the face data of the target user pre-stored in the database . Since the order of magnitude of the aforementioned face images may be very large, for example, there are 100,000. However, there may be some photos of poor quality in these face images, and the recognition accuracy is relatively low, which leads to low comparison efficiency and occupying system resources if these low-quality photos are still recognized. . [0003] In the prior art, usually a multi-task network model or a single-task network model is used alone to judge the photo quality. If only a single-task network m...

AI Technical Summary

Problems solved by technology

If only a single-task network model is used to judge, then the quality score of the training set needs to be manually marked, which has two disadvantages: 1. The efficiency of manual marking is low, which greatly affects the efficiency of model training; 2. The quality of photos needs to be considered from multiple dimensions Considering that manual labeling will be easily affected by subjective factors, making the labeling data not accurate and objective

If only the multi-task network model is used for judgment, then there are problems of large model and low operation speed

Images