Researchers are exploring various deep learning approaches for the detection and classification of diabetic retinopathy, including the use of vision transformers and convolutional neural networks. Different frameworks and models are being developed and tested, such as multi-task deep learning frameworks and robust deep ensemble CNNs, to improve accuracy in resource-constrained settings. The goal of these efforts appears to be the development of optimized systems for early diagnosis and grading of diabetic retinopathy from retinal images.