Diabetic Retinopathy Prediction Based on CNN and AlexNet Model

Diabetic retinopathy (DR) is a leading cause of vision loss in adult diabetics and keeping the vision consistent is dependent on early detection and treatment. Regular screening for these diseases is especially important for preventing progression. Convolutional neural networks (CNN) particularly, have the potential to detect and classify diabetic retinopathy from fundus images more efficiently and effectively. The objective of this study is to build a CNN-based model for detecting and categorising diabetic retinopathy using the APTOS dataset. The APTOS dataset is a sizable, openly accessible collection of fundus images that ophthalmologists have analysed for the possibility and severity of diabetic retinopathy. The accuracy obtained through the CNN model and AlexNet model is 97% and and 93 % respectively. APTOS dataset is used to train the model and a different test set is used to validate the model performance.