Detection diabetic retinopathy with the Binary Convolution Neural Network

Main Article Content

Christopher Francis Britto, Dr Divya Midhunchakkaravarthy

Abstract

                                                                                                                             


            Diabetic Retinopathy is a rapidly developing illness it can result in visual loss if not discovered and treated promptly. Early detection is advantageous since it slows the progression of disease and lowers the cost of recovery. Domain specialists play a big role in the current DR detection method. To fight this problem, we offer Binary Convolutional Neural Networks, that dramatically minimize ram use will be reduced, and the implementation will be speed up.In terms of DR classification, our technique is both hardware effective and accommodating.The Kaggle dataset was used in the experiments. Several image processing techniques, feature extraction approaches, and BCNN-based classification are used in our methodology. This design is exceptionally durable and light, and it has the ability to perform admirably in micro real-time applications with little computational capacity, allowing the screening process to be sped up. Our model was put through its paces in five different classes: No DR, Mild DR, Moderate DR, and Severe DR and Proliferative DR. When compared to the standard model, experiments utilizing the Kaggle dataset reduced memory disk consumption by 37% and increased duration by 49%.

Article Details

Section
Articles