Abstract: Diabetic Retinopathy (DR) affects 1 in 3 peoples with diabetes and remains the leading cause of blindness in working-aged adults. Currently, detecting DR is a manual processing and time consuming that needs a trained clinician and evaluate digital color fundus photographs of the retina of eye. By the time readers can submit their reviews, often a day or two later, the delayed results lead to lost follow up, miscommunication, and delayed treatment. The need for a comprehensive and automated technique of DR screening has long been accepted, and previous efforts have made good progress using image classification, pattern recognition, and machine learning. This system uses ensemble learning technique to detect diabetic retinopathy from scan images. Inception V-4, Xcpetion, ResNeXt are the three base models used which are modified and optimized to avoid overfitting, underfitting issues. The base models are individually optimized and their predictions from final layers are combined using algebraic combiner (maximum rule). Each base model differ in performance when using different datasets, therefore final output does not always depend on one higher performing base model.
Keywords: Xception, Inception V-4, ResNeXt, Ensemble Learning
| DOI: 10.17148/IJARCCE.2021.10119