Brain Tumor Segmentation Using U-Nets

In this Project a U-Net-inspired architectures were developed utilizing the concept of deep residual learning, the models were created using different building blocks and built upon different residual networks backbone architectures like ResNet50, below is the architecture of the created neural network model.

The model training and validation were done using the BraTS 2020 training dataset and testing was done using the BraTS 2020 validation dataset, below is a screenshot of the model output after testing it on one of the test cases.