Alzheimer's is a neurological condition that affects millions of individuals throughout the world. Early detection is critical for optimal treatment and management of this condition. This research presents a classification model that uses Convolutional Neural Networks (CNNs) to reliably identify the Alzheimer's disease using brain MRI scans. A publicly available dataset of brain MRI images divided into four categories: mild dementia, moderate dementia, non-dementia, and very mild dementia, has been utilized. When compared to the other classes, the number of samples for 'ModerateDemented' was found to be significantly lower, showing class imbalance. To solve this, the study oversamples the data by Synthetic Minority Over-sampling Technique (SMOTE) and generates extra samples. The proposed method augments the data utilizing the TensorFlow ImageDataGenerator. The study uses a CNN model with several sorts of normalization blocks. The suggested model achieved an accuracy of 95.2%, for the test set outperforming prior state-of-the-art techniques.