The detection and classification of the stages of Alzheimer’s disease (AD) stages through the application of the enhanced approach of convolutional neural networks and Synthetic Minority Oversampling Technique-based data augmentation. The proposed methodology addresses class imbalance and uses advanced transfer learning techniques for robust feature extraction from MRI data. Our dataset had 6400 MRI images grouped under four stages of AD. Data augmentation through rotation, zooming, and flipping augmented the generalization capability. The proposed model successfully classified the AD stage at 99% with better performance than the current model, EfficientNet -B2. This work, in fact, puts forth the potential of deep learning frameworks in the early diagnosis sense and also suggests clinical applications and further work in terms of generalizability and how multi-modal data might integrate.