<p>Detecting and classifying brain tumors accurately from MRI scans is a challenging task in medical imaging. Traditional manual diagnosis by radiologists is often time-consuming and prone to errors, making it necessary to develop automated, high-precision methods. This study presents a hybrid deep learning approach that combines two well-known pretrained convolutional neural networks—VGG16 and InceptionV3—through a late fusion strategy. The proposed model uses transfer learning, preprocessing steps such as normalization and data augmentation, and multiple regularization techniques to improve robustness and avoid overfitting. Tested on a publicly available MRI dataset containing four classes (glioma, meningioma, pituitary tumor, and no tumor), the model achieved a validation accuracy of 98.55% along with strong precision, recall, and F1-score values. Performance evaluation was conducted using a confusion matrix, classification reports, and accuracy-loss curves. Comparisons with existing models from the literature demonstrate that the hybrid architecture outperforms stand-alone CNNs in both accuracy and generalization. The results suggest that integrating complementary feature extraction capabilities of VGG16 and InceptionV3 can enhance brain tumor classification, paving the way for more reliable AI-assisted diagnostic tools in clinical practice.</p>