Brain tumors give rise to serious health threat, and timely identification is imperative for productive treatment. Manual interpretation of MRI images is subjective and can be prone to human error. The complexity and variability of tumor characteristics demand advanced diagnostic tools. In this research, we propose an advanced approach for the identification of tumors in brain utilizing Convolutional Neural Networks (CNN), an algorithm of deep learning, showcasing not only high performance through traditional training but also heightened accuracy through transfer learning. The initial CNN training phase achieved a commendable 95.31% training data accuracy and 88.44% validation data accuracy. Subsequently, transfer learning was introduced, resulting in a significant boost in performance, with training accuracy reaching 99.78% and validation accuracy at 98.64%.