This research highlights the importance of timely and precise identification of retinal diseases to prevent vision impairment and improve clinical outcomes. We apply Deep Learning (DL) techniques to classify and predict seven common eye conditions. Retinal images are critical for diagnosis, offering essential insights for ophthalmologists due to the retina’s responsiveness to changes in microvascular structures. DL has advanced medical imaging significantly, especially in tasks like classification, segmentation, detection, and prediction of retinal diseases. However, the complexity of DL models can result in decision-making errors, necessitating the estimation of model uncertainty, which traditional DL models cannot provide. Therefore, Bayesian DL methods are being increasingly used. In this study, we develop a simple Convolutional Neural Network (CNN) model for retinal disease classification and apply Bayesian CNN methods using Variational Inference (VI) and Monte Carlo dropout (MC-Dropout) to obtain the posterior predictive distribution. Using the Dataset ‘Derbi_Hackathon_Retinal _Fundus_Imag_Dataset,’ our results show that the proposed models outperform state-of-the-art models, with test accuracies of 84% for CNN, 85% for BCNN-VI, and 84% for MC-Dropout. We also compute the predictive distribution to quantify model uncertainty, demonstrating the potential benefits of Bayesian DL methods in enhancing the accuracy and reliability of disease diagnosis and treatment in medical image analysis.