Semantic Segmentation of Medical Images using UNet
Architecture & dataset used in the code: https://arxiv.org/pdf/1908.08004.pdf Training a Deep Neural Network is a trial and error process. To train a neural network for the dataset in our hand, we generally select hyper-parameters as suggested in a research paper or as provided by software libraries. Optimal values of these hyper-parameters can vary depending on learning task, dataset, weight initialization, etc. Here we will be investigating strategies to select a set of hyper-parameters.
We have tried different loss functions and regularizers and compare their performances