Analysis of Coloboma Defected Eyes using Automated Pre-trained CNN Models

Coloboma is a defect of the iris of the eye. It takes place before the birth of the child. It can affect one or both eyes. If treated from the early stages, it can restore most of that person’s vision. To detect it in the early stages, we need the help of machine learning techniques and pre-trained CNN models so that the accurate and speedy results do not take the time it usually takes otherwise. VGG16, VGG19, and Inception V3 are three feature extraction models discussed in this article that provide crucial yet simple combination techniques that can lead to more accurate predictions. We compare the above extraction models with different classifiers for the highest accuracy. After examining the ROC and Confusion Matrix, we find the best suitable model and classifier pair. In this case, we get the largest area under the curve, 95.9%, with inception V3 and neural network classifier.