Classification model training method and device

Abstract

The invention discloses a classification model training method and device, and the method and the device improve the precision of an obtained label through separately calculating the binary cross entropy of each label. In detail, the method comprises the steps of firstly acquiring a plurality of first training samples, wherein each first training sample comprises a first training image and a firstpreset number of labels corresponding to the first training image, and the first preset number of labels comprise labels corresponding to upper classifications and lower classifications correspondingto image contents respectively; then, carrying out machine learning training according to the plurality of first training samples to obtain an initial classification model; after the initial classification model is obtained, obtaining the binary cross entropy of each label in the initial classification model to serve as a sub-error value of the label; and finally, obtaining a total error value ofthe initial classification model according to the sub-error value of each label, and adjusting the initial classification model according to the total error value to obtain a target classification model.

Description

technical field [0001] The present application relates to the technical field of image processing, in particular, to a classification model training method and device. Background technique [0002] In the prior art, when classifying images, there are two common classification methods. One of the classification methods is the single-label classification method. In this method, each image only hits one label, that is, each An image is only divided into one category, so this classification method cannot fully express the semantics of the image; another classification method is multi-label classification, in which the same image can be hit with multiple labels, that is, Say, each image can be divided into multiple categories. In the existing multi-label classification algorithm, if the meaning corresponding to one label among the multiple labels corresponding to the same image is the upper classification of the meaning corresponding to another or more labels, that is, there is ...

AI Technical Summary

Problems solved by technology

Therefore, the existing multi-label classification methods have the problem of low output label accuracy

Images