Method and device for training image purification model

Abstract

The invention is to provide a method and device for training an image purification model. Compared with the prior art, the method is characterized by carrying out picture expansion on existing pictures to obtain expanded pictures; carrying out clustering on the expanded pictures to obtain corresponding clustering results; selecting preset number of pictures in at least one clustering result as sample pictures, which are preset to a user; obtaining positive and negative samples obtained after carrying out corresponding operation on the clustering results by the user; and according to the positive and negative samples selected by the user, training the corresponding image purification model, so that image quality purification can be carried out through the image purification model, and high-quality data can be obtained at a low cost. Compared with tens of thousands of and hundreds of thousands of manual labeling quantity in the past, the method allows the user to finish the labeling taskof small samples in a few minutes; model training is started to obtain the image purification model for mass image quality purification; and furthermore, more high-quality pictures can be mined fromthe mass picture data through the image purification model.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a technique for training a picture purification model. Background technique [0002] The quality purification of image data is a critical step in obtaining training data. Especially in the field of deep learning, the vast majority of methods are data-driven, resulting in the quality of image data directly related to the performance of algorithm models. Therefore, obtaining high-quality training data is an extremely important step in algorithm research. [0003] Currently, image data purification methods mainly include algorithm-based automatic mining methods and manual labeling methods. The method based on algorithmic mining is low-cost, but the effect cannot be guaranteed. Although the manual labeling method is of high quality, it is costly and takes a long time, especially for massive data, which often has tens of millions or even hundreds of millions of records. Th...

AI Technical Summary

Problems solved by technology

The method based on algorithmic mining is low-cost, but the effect cannot be guaranteed

Although the manual labeling method is of high quality, it is costly and takes a long time, especially for massive data, which often has tens of millions or even hundreds of millions of records. The manual labeling method cannot well meet the business needs of big data.

Images