<p>Diabetic retinopathy (DR) is a perilous ocular ailment that impacts individuals with diabetes. Diabetes impacts more than 60 million individuals in India. By 2030, this pervasive problem is anticipated to have increased to approximately 578 million cases. The individual continues without any symptoms related to the condition of DR until their eyesight is impacted. The effectiveness of treatment is most important when administered before the advancement of the illness. Hence, timely detection of DR is essential for its management, since it has the potential to lead to irreversible vision loss. Diabetic retinopathy (DR) is the prevailing consequence of diabetes, affecting around 3 to 4.5 million individuals in India with the potential to cause visual impairment. The therapy necessitates expensive equipment and pharmaceuticals, and the condition needs consistent monitoring from the initial prognosis until death. Furthermore, the process of visually examining fundus pictures by skilled ophthalmologists to detect structural alterations in microaneurysms, exudates, blood vessels, hemorrhages, and the macula is an extremely laborious task. It is also prone to significant fluctuation in observations made by different observers and within observations made by the same observer. Several contemporary deep learning methods are presently used to classify input images automatically. This research presents an exhaustive examination of deep learning methodologies implemented in retinal image analysis to detect and classify diabetic retinopathy (DR). The variables that may impact the efficacy of a deep learning system in identifying diabetic retinopathy (DR) are also taken into account.</p>