Eye diseases are crucial to be diagnosed at the initial stage. The physical diagnosis is inaccurate, expensive, slow, and could not be available to people in inaccessible or underdeveloped locations. High-resolution photographs of the eye's internal components can be obtained using imaging techniques like OCT and fundus photography, but these methods also necessitate expensive equipment, trained operators, and may not be able to spot early-stage illnesses. Additionally, these techniques provide a lot of data that can be challenging to analyse and interpret, delaying both diagnosis and therapy. Due to the shortcomings of present techniques, a more precise and effective strategy for finding eye illnesses is required (Rafay A et al., 2023). Machine-learning algorithms provide faster responses and are more accurate. However, the inherent limitations of machine learning algorithms result in compromised accuracy, which can be improved using the latest deep learning algorithms. In this context, the proposed study uses a MobileNet convolutional neural network, which has superior performance in detecting cataracts, glaucoma, and diabetic retinopathy compared to normal eyes. The results indicate that the proposed algorithm results in 95% accuracy in detecting the mentioned diseases, which is the highest in contrast to recent state-of-the-art algorithms observed in the existing literature.