Multi-Class Classification of Brain Tumours: Leveraging VGG, InceptionV3, and DenseNet201 Transfer Learning

In this study, brain tumour is classified on the Brain MRI(bt_dataset) dataset using Visual Geometry Group16(VGG), Inception V3 (Inv3), and DenseNet201(DN201) deep learning architectures. In this process, the models were developed for each architecture using Transfer learning, results were achieved, and the model with a minimum loss was introduced. The dataset used for the study is taken from the Kaggle website. The classification process was applied repeatedly on the bt_dataset, and the multi-class classification or prediction with the minimum error rate was aimed at the established models. In the study process, initially the bt_dataset was collected, prepared, pre-processed and visualized. Models were developed using the Jupiter Notebook editor via the Google Colaboratory. Then, the models were further examined and the most successful architecture was selected. As a result, according to the prediction/classification models, it was observed that the VGG architecture has achieved better results than the Inv3 model and the DN architecture.