Label noise-containing image classification method based on graph consistency and semi-supervised model

The invention relates to the technical field of computer vision and artificial intelligence, in particular to a label noise-containing image classification method based on graph consistency and a semi-supervised model. The method comprises the following steps: S1, training total noise label data in a training set, and initializing a model; s2, obtaining the distribution of samples; s3, classifying and screening the samples; s4, performing different data enhancement on the image; s5, the enhanced image is sent to a semi-supervised model for training; s6, performing graph coding to obtain a consistency graph; s7, jointly optimizing the model, and updating a sample label; and S8, taking the semi-supervised model as an inference model. The method mainly solves the problems that an existing semi-supervised method only based on classification consistency training faces a very serious confidence deviation problem, a large number of noise labels are remembered, errors are accumulated, and a model is damaged. The grap