Accurate disease diagnosis is a primary goal of medical image analysis. This work provides a novel algorithm that is specifically designed to identify retinal illnesses inside medical photographs. It is based on YOLOv5 (You Only Look Once version 5). The suggested method makes use of YOLOv5's strong object detection capabilities to provide remarkably effective real-time end-to-end illness detection. By utilizing the carefully selected Eye Net dataset, our model is subjected to extensive training and optimization, which leads to remarkable ability in identifying various retinal disorders such as glaucoma, macular degeneration, and diabetic retinopathy.
An extensive analysis clearly demonstrates this approach's superiority over traditional Convolutional Neural Networks (CNN) and other models. The suggested model outperforms the benchmark models with U-Net segmentation, Deep Learning CNN, Shuffle Net and SVM, multi-class SVM, and DCNN, which reach accuracies ranging from 89.3% to 95%. The new model achieves an impressive accuracy of 98%. This achievement represents a significant advancement in the area and holds the potential to improve patient care by facilitating prompt interventions and therapies, raising the standard for the identification and categorization of retinal disorders. As a result, the suggested strategy stands out as the pinnacle of precision and effectiveness, opening the door for revolutionary developments in the medical industry.
Keywords: YOLOv5, Eye Net, Retinal illness, Ophthalmology, Convolutional Neural Networks.
How to cite this article: Tata R, Rajesh Y. A novel approach multi-class retinal disease detection with YOLOv5: accurate and real-time diagnosis. Int J Drug Deliv Technol. 2026;16(3s): 1025-1031; DOI: 10.25258/ijddt.16.3s.123
Source of support: None.
Conflict of interest: None