Diabetic retinopathy (DR) is a significant consequence of diabetes that affects the blood vessels in the retina, potentially leading to visual impairment or blindness. Traditional diagnosis entails a physical examination of fundus images, which is prone to inaccuracy. In response, researchers advocate for the use of computer vision techniques to automate DR detection. However, these approaches frequently meet difficulties, notably in reliably diagnosing early-stage DR. This research digs especially into the use of Machine Learning (ML), adopting a DenseNet-121 architecture for the analysis of DR stages. A significant dataset from the APTOS 2019 Blindness Detection competition on Kaggle serves as the basis for this exploration. The method uses patient fundus images as input, allowing the model to extract features and generate output via an activation function. With a stated accuracy of 0.9644 and a quadratic weighted kappa score of 0.9135 for DR detection, remarkable results are obtained. This highlights the revolutionary potential of AI and ML in the diagnosis and management of diabetic retinopathy. It opens the door to more accurate and efficient screening and treatment approaches.