Diabetic retinopathy can cause visual disability; timely identification of diabetic retinopathy is essential to avoid irreversible loss of eyesight. Deep learning is powerful in automation of diagnosis process and categorizing diabetic retinopathy from retinal images. This paper provides detailed overview of the latest developments in the area of machine learning-based diabetic retinopathy identification. We examine the difficulties posed by human screening techniques, the shortcomings of conventional diagnostic strategies, and the possibilities presented by machine learning algorithms for improving the effectiveness and precision of screening. The pros and drawbacks of the different deep learning and machine learning models used to identify diabetic retinopathy are explored. In addition, we look at how feature extraction strategies, preprocessing approaches, and performance evaluation criteria all contribute to improving model performance. We also discuss the application of machine learning models in clinical practice. With this review, we discussed diabetic retinopathy detection using machine learning. The CNN model performs well across a range of evaluation criteria, according to its overall performance metrics. With a 91% accuracy rate, the model showed that it could accurately classify occurrences of diabetic retinopathy.