Eyes on Health: A Deep Learning Approach for Ocular or Eyes Disease Detection

At present, eye conditions or eye diseases are the primary reason for blindness and vision loss around the globe. Identifying and forecasting eye illnesses with conventional techniques are difficult, lengthy, and susceptible to human mistakes. Any delay in identifying and foreseeing future issues will lead to the patient losing their eyesight or experiencing vision problems. In order to prevent this, we require a system that is capable of detecting and predicting eye diseases with effectiveness, efficiency, and automation. In this article, we developed a computerized system using a pre-existing EfficientNetB3 CNN family structure to detect eye diseases through deep learning technology. The dataset utilized in this research was obtained at no cost from the web and comprises 4,217 pictures. The model was trained with transfer learning and enhanced with an improved dataset through augmentation methods. Precision score, Recall score, and F1-score and accuracy are utilized for assessing the model’s performance. Our findings indicate that the EfficientNetB3 model performs better than conventional CNN structures, achieving high precision in identifying various eye diseases. The accuracy of model 97% on the training dataset and 95% on the test dataset by incorporating image augmentation. We utilized characteristics from the EfficientNetB3 structure, tailored for identification and categorization in image datasets.