Fundus image classification method and device and computer readable storage medium

Abstract

The embodiment of the invention provides a fundus image classification method and device and a computer readable storage medium. The fundus image classification method comprises the steps of obtainingfundus images and categories of objects included in the fundus images; adjusting a display parameter of the fundus image to meet a value condition; converting the display parameters of the fundus image according to the distribution conditions required to be met by the distribution of the display parameters to obtain a new image; training a neural network model based on a training set constructedby the adjusted fundus image and the new image and the category of the included object; and based on the trained neural network model, determining the category of the object included in the fundus image to be predicted.

Description

technical field [0001] The invention relates to image processing technology, in particular to a fundus image classification method, device and computer-readable storage medium. Background technique [0002] With the wide application of multimedia technology and computer equipment network, a large amount of image data appears on the network. How to effectively manage these image files, and automatically identify and classify the content of these images becomes more and more important. [0003] At present, with the continuous improvement and development of machine learning methods, more and more attention has been paid to deep learning algorithms. Among them, convolutional neural network is an important algorithm in deep learning, which has become a research hotspot in the field of image recognition. Image classification technology based on convolutional neural network can automatically extract feature information from images, and express images through the extracted features...

AI Technical Summary

Problems solved by technology

[0004] However, when classifying images based on convolutional neural networks for different specific fields or categories, it is often necessary to establish network models corresponding to the fields or categories, and the layers and overall architecture of each network model are trained. To be sure, in order to obtain better feature expression ability to achieve better classification accuracy, especially for fields that require higher classification accuracy, it is often necessary to obtain more identical or similar original image data for images in the same field or category As training data, in order to increase the depth of the network and expand the network scale, the structure of each level of the network model is determined based on the structure of the previous level through more training data, and then constructed, so that the training method of the network model is complicated, and the training is obtained. The classification accuracy of the network model of the image is not stable enough

Images