Published Paper Details:

Using the VGG16 model and transfer learning, this algorithm classifies retinal pictures as either exhibiting indications of diabetic retinopathy ("dr") or not ("nodr"). Images are processed with ImageDataGenerator for augmentation using TensorFlow and Keras, and then loaded in batches for testing, validation, and training. Only the last layer of the VGG16 model is altered for binary classification,the other layers are loaded with pre-trained weights. A new dense output layer with softmax activation is added after the other layers have been frozen to preserve learnt features. The Adam optimiser and categorical cross-entropy loss are used to train the model over five epochs.Lastly, a confusion matrix is used to visualise the test set predictions, emphasising true positives and false negatives to assess classification performance. Using transfer learning, Confusion Matrix method uses little data to produce accurate, efficient classification