Implementasi CNN terhadap Sistem Klasifikasi Penyakit pada Retina Mata Manusia Menggunakan Model MobileNetV3 Berbasis Website

Abstract

Eye is one of the most important organs for humans where the eye can obtain almost 80% of information only through the sense of sight, if the eye experiences problems in its function, then the eye cannot see the object in front of it clearly or perfectly. This will be a problem for humans because the eye is a very crucial organ in carrying out any daily activities. If eye diseases are not identified and treated early, the condition of the eyes may become worse and may even cause blindness. Therefore, a way is needed so that identification of the eye can be done more quickly and accurately, eye diseases can be identified through the retinal fundus of the eye. One solution that can be used is Machine Learning technology by relying on the Convolutional Neural Network (CNN) algorithm from the Deep Learning discipline. CNN is one of the algorithms in Machine Learning concept that can process complex data such as images. With the development of Machine Learning science, CNN continues to be developed so as to produce a collection of faster models that have been trained as pre-trained models. One form of pre-trained model is MobileNetV3 which is the latest version of the previous model version, MobileNetV2. CNN along with MobileNetV3 will be used to identify the type of condition that the eye has through the fundus given. The accuracy results obtained are 98% for training data and 90% for validation data, the implementation will be used in a web-based application.