Eye Diseases Classification Using Transfer Learning of Residual Neural Network
Yubo Cui, Ruo‐Yao Sun, Boyang Liu, Zhiyang Liu, Teoh Teik Toe
Abstract
Eye diseases have long been a serious problem for people, including glaucoma, diabetic retinopathy, cataract and so on. Fortunately, their images in the fundus are distinguishable, which makes it possible to distinguish eye diseases by deep learning. In this paper, we propose an eye disease classification method based on transfer learning, in which ResN et model is used. After the fundus retinal image data is processed and the model structure is optimized, the accuracy on the test set reach 0.92908, and the accuracy on the training set is close to 1. We then calculate some indicators to evaluate our model. Meanwhile, we also use other pretraining models to replace ResN et for experiments, and finally find that our model has excellent performance in distinguishing different eye diseases.