Diabetic eye disease is a complexity that affects people having diabetes for a longer time. By affecting the blood vessels it can cause blurry vision or even blindness to the patients. Thus, detecting eye disease at an early stage can help many diabetic patients to get the required treatment and intern increases the survival rate. In the proposed system, the CNN algorithm of machine learning is used to detect diabetic eye disease by using thermal images. These images are pre-processed by converting them from RGB to GRAY based on which the required features are extracted. To detect diabetic retinopathy, here the Convolutional Neural Network is used to classify 5 stages of the diseased eye.