Fundus Retina Image Classification Method Based on Capsule Network

Abstract

The invention discloses a method for classifying fundus retinal images based on a capsule network, which belongs to the technical field of image processing. In order to achieve the purpose of improving the classification accuracy of fundus retinal images and reducing the training time of classification models, the technical solution adopted in the present invention is: firstly construct a typical capsule neural network, and pre-train it on the fundus retinal images to obtain the capsule neural network. The pre-training weight of the neural network; then obtain the pre-trained AlexNet model weight parameters from the public data, replace the last two fully connected layers with the trained capsule neural network layer, call the network structure CapsAlexnet, and then CapsAlexNet After training and fine-tuning on the fundus retinal images, the fundus retinal classification model with the best classification accuracy and fast convergence is finally obtained, and the accurate classification processing of the fundus retinal images is realized.

Description

technical field [0001] The invention belongs to the field of image processing, in particular to a fundus retinal image classification method using a capsule network, and belongs to the technical field of biomedical image classification. Background technique [0002] With the rapid research and development of deep learning in image classification and recognition in recent years, it has also received more and more attention in the field of biomedical imaging, and has gradually had many implementation scenarios and reliable applications. Important applications in many fields, such as etiology analysis, disease detection, and telemedicine, are gaining more and more attention just like the explosion of deep learning. [0003] In recent years, due to the rapid progress in the field of deep learning, there has been tremendous development in the automatic classification and semantic annotation of images. It can be said with certainty that in recent years, deep learning has dominate...

AI Technical Summary

Problems solved by technology

However, due to the complex structure of fundus retinal images, and the lack of obvious differences and specificities between diseases of the same disease, it is difficult to apply the automatic processing of fundus retinal images to practical scenarios.

Images