Prolonged hyperglycemia and blood glucose fluctuations are the main causes of diabetic retinopathy (DR), one of the primary causes of visual impairment worldwide. For working-age people, prompt intervention is essential to maintaining vision. This study is driven by the urgent need for precise and early DR identification in order to enable efficient treatment and avoid permanent vision loss. Based mostly on retinal fundus photos, we present a thorough analysis of a number of diagnostic indicators, such as blood vessel anomalies, microaneurysms, exudates, macula, optic discs, and hemorrhages. Our contribution consists on assessing several AI approaches, including deep learning and machine learning, for early DR classification and detection. This study specifically demonstrates that a hybrid model that combines a Support Vector Machine (SVM) classifier with ResNet50 for feature extraction yields improved accuracy in detecting diabetic retinopathy.