Brain tumors are characterized as an abnormal mass within the brain tissue, and they can either be malignant or benign. Malignant tumors are aggressive and grow rapidly, while benign tumors, even though they are not harmful, may exert pressure on important parts of the brain and can still pose a threat. Advanced imaging methods such as magnetic resonance imaging (MRI) are essential in making accurate diagnoses without surgical intervention. This study presents a new technique by which brain tumors can be automatically detected and classified in MRI images to this end, a convolutional neural network (CNN) model has been proposed for binary and multiclass classification of brain tumor. Performance enhancing approaches such as data augmentation and dimensionality reduction are also performed as part of the model. We apply a deep CNN processing each slice of the brain MRI into many layers with ReLU activations and max-pooling layers for feature extraction. The model determines during training, an accuracy of 93% while during testing; an accuracy of 90% is achieved. Transfer learning is applied in this work to mitigate the problem of data scarcity being a source of improvement in diagnosis accuracy using small datasets. The propounded framework shows that brain tumor diagnosis utilizing deep learning models can be successfully integrated into routine clinical diagnosis.