Tumors in the brain are the uncontrolled development of cells. This uncontrolled development may occur in various regions of the brain, affecting the proper functioning of the brain, leading to seizures, paralysis, loss of hearing, loss of vision, and death. If diagnosed early, it can reduce the toll due to brain tumors. With the advancement in deep learning, CNN (Convolutional Neural Network) models are proven to be trailblazing performers for image classification, segmentation, and object detection. This development has aided the widespread use of CNN architectures aimed at medical picture categorization. Transfer Learning approach is employed for categorizing picture collections from Magnetic Resonance Imaging (MRI) for this study. The EfficientNetV2-L architecture is used for fetching the patterns and fine-tuning with a slow learning rate. The formulated architecture has been trained using various hyperparameter settings. The formulated architecture, with 512 Neurons in the dense layer, shows good validation accuracy for brain tumors categorization into four labels: glioma, pituitary, normal, and meningioma.