Detection of eye defect with the aid of answerable AI

A cataract is an issue with the eye’s typically clear lens. Ocular abnormalities that cause vision blindness are frequently seen. Early detection of the cataract will reduce danger and shield you from becoming blind. In this paper, proposing a model to identify defect in an eye using a supervised machine learning algorithm. Here, A black box and white box technique is used. A black box technique used to classify the image while white box technique used to explain the output given by black box technique. Using this method, a user clearly understands the reason behind the output. In black box method, CNN model is used. It helps to classify the input image into cataract image or non-cataract image. In white box method, SHAP algorithm is used, which shows the part of the eye having cataract. The input to the model is a fundus image collected from high resolution camera. The result of proposed CNN model is compared with VGG16, ResNet50, InceptionV3. The proposed CNN model give accuracy 94%.