Cataract Detection using Hybrid CNN Model on Retinal Fundus Images

Abstract

In this research, a hybrid convolutional neuron network (CNN) model was developed for cataract detection. The full fundus image in the original dataset will be divided into four segments that created five fundus image datasets and trained by five different CNN models which have the same structure. The five model predictions will pass through majority voting to get the final prediction. The experimental result shows that the proposed hybrid CNN performs better than stand-alone models.