Diabetic retinopathy (DR) is a serious condition that results from high blood sugar’s effects on the eye. The damage done to the blood vessels (BV) in the light-sensitive tissue at the back of the eye causes this disorder. Interestingly, it is a frequent cause of blindness in people in their prime working years, and the risk is increased when diabetes is not adequately managed. The BF-CNN model has been significantly improved in this ground-breaking study to strengthen its ability to identify diabetic retinopathy. The suggested model deliberately uses cutting-edge methods for both feature set selection and classification. By adding new criteria based on biological processes, such as chemotaxis, reproduction, elimination, and dispersal, the model is better able to identify complex patterns that are suggestive of diabetic retinopathy. This enhancement helps to identify the disease in retinal pictures in a more accurate and subtle manner. Additionally, the model makes use of the convolutional neural network (CNN) architecture, a complex family of artificial neural networks distinguished by fully-connected, pooling, and convolutional layers. A thorough examination of the intricate visual information present in retinal pictures is made possible by this technological combination. The study model has been trained on a sizable dataset of 3000 publicly accessible retina photos that were selected by eyepatch on the KAGGLE platform through the integration of high-performance computing resources, such as GPU acceleration. The results of this effort are encouraging, as the suggested model has an amazing accuracy rate of almost 98.01%. The study uses extensive performance criteria, like as accuracy, precision, recall, and more, to evaluate its effectiveness. This study offers important implications for early management and better patient outcomes by utilizing state-of-the-art technologies to increase the precision and dependability of diabetic retinopathy identification.