An extensive comparative examination of Diabetic Retinopathy (DR) detection methods is presented in this paper, with a focus on examining the efficacy of deep learning methodologies and segmentation strategies, primarily Convolutional Neural Networks (CNNs) and Residual Networks (ResNets). The work starts with a thorough literature analysis that covers the most recent developments in deep learning architectures and conventional segmentation techniques. CNNs and ResNets are particularly well-suited to capturing complex retinal properties that suggest deep learning. Using publicly available retinal image datasets, a range of segmentation techniques, CNNs, and ResNets are carefully tested during the project’s implementation phase. Evaluation metrics are used to thoroughly evaluate each methodology’s performance, including sensitivity, specificity, accuracy, and other pertinent metrics. In addition, the research explores how the size, diversity, and picture quality of the datasets affect the effectiveness of the suggested techniques, offering insights into the findings’ generalizability and practical use. The results of this thorough comparison study are intended to provide detailed insights into the advantages and disadvantages of segmentation techniques, CNNs, and ResNets for DR detection. The paper’s results help practitioners and researchers create earlier diagnosis strategies for DR that are more successful. This work promotes the development of precise and understandable tools, which eventually improve the outcomes of diabetic eye care and open the door for improved patient management techniques.