Diabetic retinopathy is an eye disorder that can impact a person's vision due to the irregular development of blood vessels in the retina. Millions of people, especially the elderly, experience this condition every year. It is estimated that around 33% of individuals with diabetes experience diabetic retinopathy to some extent and as the prevalence of diabetes continues to rise, the prevalence of diabetic retinopathy is also expected to increase. Detection of diabetic retinopathy at an early stage is crucial because it is irreversible and can lead to blindness if left untreated. The traditional method of diagnosing diabetic retinopathy by ophthalmologists is a time-consuming process and may result in incorrect diagnosis due to the possibility of human error. The advancements made in deep learning (DL), particularly in convolutional neural networks (CNN), have significantly improved fundus image analysis. In this study, we calculated the accuracy of 5 different types of CNN architecture; compared models based on accuracy, and predicted diabetic retinopathy in new samples of fundus images using the most accurate model.