For better patient outcomes and treatment effectiveness, brain tumors must be detected early. The efficiency of Convolutional Neural Networks (CNN) and Transfer Learning using MobileNetV2 in detecting brain cancers from MRI data is compared in this study. Models were trained and tested using the MRI scan dataset to guarantee compu- tational efficiency and classification accuracy. The accuracy of the sug- gested model’s categorization was 99.36%. The results indicate that both techniques achieved excellent accuracy in tumor classification, with CNN marginally exceeding MobileNetV2. MobileNetV2’s pre-trained weights resulted in faster training times and reduced computing complexity. These findings indicate that MobileNetV2 is a feasible choice for resource- constrained situations, whereas CNN may provide better performance. More study is needed to investigate hybrid techniques and adjust model parameters for greater accuracy and efficiency in clinical situations.