This paper addresses the classification and prediction of eye diseases, specifically Glaucoma, Diabetic Retinopathy, Cataract, and normal cases, utilizing advanced deep learning techniques. The primary objective is to assess the performance of the ResNet18 model, enhanced with transfer learning and various optimizer methods, in accurately identifying these conditions. The dataset used in this study consists of 1757 retinal fundus images, including 453 normal cases, 441 cases of Diabetic Retinopathy, 476 cases of Glaucoma, and 387 cases of Cataract. The images were resized to 224x224 pixels and normalized for training. The ResNet18 model, pre-trained on the ImageNet dataset, is fine-tuned through transfer learning to specialize in eye disease classification. Three optimization methods—SGDM, ADAM, and RMSProp—are employed to optimize the learning process. Cross-validation, specifically a 5-fold method, is utilized during training to enhance the model's reliability and robustness. The performance metrics, which include accuracy, precision, recall, and the Area Under the Curve (AUC), demonstrate a high accuracy rate of 91.4% and an exceptional AUC value of 99.99%. While the results confirm the effectiveness of the model in identifying various eye diseases, there were challenges such as overfitting and the need for further with external data. Future research could explore the performance of the model on larger, independent datasets to validate its generalization capabilities and expand the range of eye diseases considered............ Keywords:........ Deep learning, Eye diseases, Glaucoma, Diabetic Retinopathy, Cataract, transfer learning, ResNet18, SGDM, ADAM, RMSProp, classification, cross-validation.