Brain tumors pose a serious threat to global health, making accurate and early detection essential for effective treatment planning. While Magnetic Resonance Imaging (MRI) is widely used for diagnosis, manual interpretation is time-consuming and subject to error. This has prompted increasing use of deep learning for automated tumor classification. We propose a novel framework based on the Swin Transformer V2 architecture for classifying brain tumors in MRI scans into glioma, meningioma, pituitary tumor, and non-tumor categories. The design introduces two core innovations: a Dual-Branch Down-sampling module and an Enhanced Attention Mechanism, which improve multi-scale feature representation... More >