Two-stage classification of glioma subtypes using a hybrid CNN-ViT model

Abstract

This repository contains the research code for glioma subtype classification using a hybrid Convolutional Neural Network (CNN) and Vision Transformer (ViT) framework. The project investigates the use of CNN-based feature extraction and transformer-based visual representations for glioma classification. In particular, the repository includes implementations based on MobileNet, MobileNet-ViT-B/16, Vision Transformer (ViT-B/16), Swin Transformer Tiny, VGG16, and NASNetMobile. The repository provides implementations for single-stage, two-stage, binary, multi-class, and ensemble classification experiments. The main objective of this repository is to support research reproducibility, model evaluation, comparative experimentation, and further development of deep learning approaches for glioma subtype classification.