Data annotation method and device and disease classification model training method

Abstract

The invention provides a method for carrying out data annotation on a sample data set, which comprises the following steps: S1, acquiring the sample data set, each sample in the sample data set comprising one or more classification labels which are respectively annotated by a plurality of annotators; s2, merging the label types of the samples containing various classification labels so as to oppositely combine the associated classification labels, and taking one label in the label pair as a merged label; wherein the associated classification label pairs refer to paired combinations formed by different labels labeled by different labeling persons for the same sample; and S3, re-labeling the samples in the sample data set based on the merged classification labels. Compared with the prior art, the method provided by the invention has the advantages that the data with certain subjectivity can be preprocessed, so that subjectivity evaluation is objectified by using other indexes with correlation to obtain a label with universality so as to label the data, and then a correlation classification model is trained.

Description

technical field [0001] The present invention relates to the field of artificial intelligence, specifically, to the field of supervised machine learning in the field of artificial intelligence, and more specifically, to a data labeling method and model for data multi-classification problems with partial subjectivity, and a method based on A disease classification model training method for fundus images and a disease classification method based on fundus images. Background technique [0002] Supervised machine learning in the field of artificial intelligence means that the machine uses the training samples of existing label information for training, compares the output of the model mapping with the label information of the training samples, and can use the existing information to train and correct the model in iterations . In supervised learning, typical problems can be divided into regression problems, classification problems and labeling problems according to the characteri...

AI Technical Summary

Problems solved by technology

It can be seen from this that for classification problems with certain subjectivity in the identification content or evaluation class, different people may get different results on the same sample, resulting in differences, so that the classification problems of supervised learning cannot be obtained. Universal label information under objective evaluation makes model training difficult

Images