In the modern world, diabetes is one of the major health problems that affect humans all around the world. Lack of early detection, prolonged diabetics, might lead to medical complications such as heart problems, eye vision problems, skin issues etc. Diabetic retinopathy (DR) is a frequent abnormality of diabetics. To help patients with the early detection of diabetic retinopathy, in this paper, we propose a computer vision-based technique to analyze and predict diabetes from the retinal input images. Convolutional neural networks (CNN) and support vector machine (SVM) are trained with diabetic and non-diabetic retinal images. Results show CNN reports better accuracy in DR compared to SVM. © 2022 Elsevier B.V., All rights reserved.