Detection of Diabetic Retinopathy using Convolutional Neural Networks

The main cause of vision loss worldwide is diabetic retinopathy (DR). This project presents a frame work for automatic DR detection using retina and retinal images. To improve overall modeling and prevent overshoot, the system uses regularization techniques and VGG16 ,CNN trained on ImageNet. Binary cross- entropy loss function trains the system for for the project. More than 90% of tests pass when ADAM optimizer is used. Achieve accuracy when developing training programs. There is a short delay to maintain the lustiness of the model. This approach will be taken to use in real clinical settings. It is a authentic instrument for early treatment.