Small sample attention mechanism parallel twinning method for eye fundus image classification

The invention provides a small sample attention mechanism parallel twinning method for eye fundus image classification, which performs classification according to an eye fundus lesion image of a patient to obtain a classification result, and comprises the following steps: reading a medical eye fundus image data set for preprocessing to obtain preprocessed image data; through a few-shot learning method based on a twinning network Siamese, a dense connection network densenet pre-trained by using a data set Image Net is migrated by using a feature-based transfer learning method to extract the features of two different images, and a convolutional block attention module (CBAM) is added on the basis of the network to select more key image information, so that the feature extraction of the two different images is realized. And similarity measurement of the pictures is carried out through a contrast loss function, so that a more accurate classification prediction result is obtained. The method migrates a dense connec