Diabetic Retinopathy (DR), a common complication of diabetes, poses a significant threat to vision, potentially leading to irreversible vision loss. Due to the irreversible nature of DR, treatment focuses on preserving existing sight, highlighting the critical need for early detection and management to prevent severe consequences. Traditional manual analysis of retinal photographs by ophthalmologists, while time-consuming and variable, is prone to false negatives compared to computer- aided diagnosis systems. The difficulty in accurately interpreting retinal fundus images by many medical professionals further complicates treatment, necessitating a lengthy, expensive, and time-intensive process. Deep learning, particularly CNN (CNN), has emerged as a promising approach, demonstrating success in medical imaging, including DR detection. This article presents a comprehensive review of state-of-the-art deep learning techniques for automated DR detection and classification using color fundus images. Furthermore, it explores available color-fundus retina datasets for DR and discusses challenging issues requiring further investigation. This article also provides an overview of recent advances in DR diagnosis and assessment using updated CNN-based techniques, reviews missed datasets in public repositories, and identifies future directions to enhance the efficiency and effectiveness of automated DR screening. Given that early diagnosis is crucial for effective treatment in the medical field, this holds true for DR, a condition linked to diabetes, affecting 425 million adults worldwide, and impacting the retina, heart, nerves, and kidneys.