In recent years, myopia (commonly known as short-sightedness) has become one of the most rapidly increasing vision-related disorders across the globe, particularly among children and young adults. This sharp rise is strongly linked to modern lifestyle patterns, including excessive screen time due to smartphones, laptops, and other digital devices, reduced outdoor activities, and prolonged engagement in near-vision tasks such as reading and gaming. As societies continue to become more technology-driven, the prevalence of myopia is expected to grow even further, making it a significant public health concern in the coming years. Despite being a common condition, myopia can lead to serious complications if not detected and managed at an early stage. High or progressive myopia increases the risk of severe eye disorders such as retinal detachment, glaucoma, cataracts, and myopic macular degeneration, which may ultimately result in irreversible vision loss. Therefore, early diagnosis and timely intervention are essential to control its progression and prevent long-term damage. However, conventional diagnostic approaches typically require specialized ophthalmic equipment and expert medical professionals, which may not always be accessible or affordable, especially in rural and underserved areas. To address these limitations, this study focuses on the development of an intelligent, automated, and efficient deep learning-based system for the early detection of myopia using retinal fundus images. Fundus imaging provides a detailed view of the internal structures of the eye, including the retina, optic disc, and blood vessels, making it a valuable tool for detecting various ocular conditions. By integrating artificial intelligence with medical imaging, the proposed system aims to assist healthcare professionals in making faster, more accurate, and reliable diagnostic decisions.The proposed approach utilizes advanced deep learning architectures, specifically ResNet50 and MaxViT, for image classification. ResNet50, a widely used convolutional neural network, is known for its ability to handle deep architectures efficiently through residual learning, while MaxViT (Multi-Axis Vision Transformer) combines the strengths of convolutional operations and transformer-based attention mechanisms to capture both local and global features effectively. These models are trained and evaluated on a large dataset consisting of more than 20,000 retinal fundus images, which are carefully divided into training, validation, and testing sets to ensure robust performance evaluation and avoid overfitting. Extensive experimental analysis demonstrates that both models perform effectively in detecting myopia; however, the MaxViT model consistently outperforms ResNet50 in terms of classification accuracy, feature extraction capability, and overall reliability. The improved performance of MaxViT can be attributed to its hybrid architecture, which enables better understanding of complex patterns in retinal images. The results highlight the potential of transformer-based models in advancing medical image analysis and improving diagnostic accuracy.