Cataract and glaucoma are the leading causes of vision impairment worldwide,according to data from the World Health Organization. In Indonesia, theseconditions rank first in Southeast Asia and second globally, as evidenced bydata from the Ministry of Health's Roadmap of Visual Impairment ControlProgram in Indonesia 2017-2030. Early detection of these diseases is crucialfor preventing blindness. This study aims to classify eye diseases using a native-architecture Convolutional Neural Network (CNN) classification method withthe novel inclusion of three non-fundus or real-eye image subsets. The CNNimplementation in this study employs 100 epochs and achieves an accuracy of98.67%. The saved model from this research will be deployed usingTensorFlow.js, a framework or library derived from TensorFlow.