Published Paper Details:

Diabetic retinopathy, caused by damage to the retina from diabetes mellitus, is diagnosed through coloured fundus images that require trained clinicians to identify various small characteristics, making it a time-consuming task. In this research, we present a CNN-based method to detect diabetic retinopathy in fundus images. Our novel segmentation strategy using Gabor filters prepares the data for training the model, and data augmentation is used to gather enough data due to the limited dataset. Our segmentation approach identifies intricate properties in fundus images and also detects the presence of diabetic retinopathy. We utilize the ResNet50 algorithm for training the model, which is efficiently trained using a high-end Graphics Processor Unit (GPU). Our proposed method shows promising results in the detection of diabetic retinopathy, indicating its potential for use in clinical practice. On training the dataset we achieved an accuracy of 71.51 percent.