Hybrid Deep Learning Ensemble Model Using ResNet50, VGG16, and MobileNetV2 for Eye Disease Identification

Abstract

Diabetic eye illnesses, including retinopathy due to diabetes and macular edema, have been a category of ocular disorders that may impact individuals. Timely identification is essential for preventing blindness and vision impairment; nevertheless, physical diagnosis is labor-intensive and requires specialized expertise. This paper introduces "Hybrid Deep Diabetic," an ensemble combination of models deep learning techniques system for the automatic diagnosis of eye illness. This system combinesEyeNet, ResNet50, VGG16, &MobileNetV2 to hybrid models that enhance classification accuracy. EyeNet used as a  bespokeDeep Learning Technique convolutional neural network (CNN) optimized in retinal imagegeneration , while models that have been trained (ResNet50, VGG16, MobileNetV2) use transfer learning to extract robust features.The hybrid model integrates predictions using an ensemble methodology, enhancing diagnostic accuracy compared to singular models. The Calculate metrics: accuracy,recall, precision, &F1-score—indicate that combined approach surpasses both solo CNNs and conventional machine learning methods. The methodelevated categorization precision, facilitating prompt identification anddiminishing reliance on human screening. Future improvements may include real-time use in clinical settings and interaction with telemedicine systems for broader accessibility.