Image classification method and device based on self-supervised active learning, equipment and storage medium

The invention relates to an image classification method based on self-supervised active learning, and the method comprises the steps: training an initial neural network model through employing a labeled first image data set and an unlabeled second image data set through the self-supervised learning, and carrying out the fine adjustment of an initial classification network model through employing the labeled first image data set; obtaining a value image data set from the unlabeled second image data set by using information entropy and self-supervision loss; and marking the value image data set and updating the first image data set, and performing fine adjustment on the optimized first classification network model again. According to the method, the neural network model is trained by utilizing self-supervised active learning, the most valuable image data is selected for labeling in combination with the self-supervised loss of a mask auto-encoder and the uncertainty of an image data sample, and meanwhile, the op