Diabetic retinopathy is a complex eye disease caused by diabetes. This causes damage to blood vessels and creates abnormal new vessels. These lead to vision loss at the end if not diagnosed properly and not treated. Our approach includes data collection, data pre-processing to remove Diabetic retinopathy images from collected data sets and standardize them. Feature extraction is done using segmentation architecture i.e., UNet to propose a segmentation strategy that separates structure and texture components and also improves the performance. Then DR fundus images are trained and then the trained DR images are classified as DR or No DR using DL neural network like Convolutional Neural Network (CNN).