Generally, training a neural network is a very tedious task as it has many hyper-parameters to tune which takes a lot of time. In this assignment we will try to find out strategies to tune one (or a few) hyper-parameters for image segmentation tasks. Our experiments include different optimizers, learning rate schedulers at different training-validation splits along with different regularizers, loss functions and different types of augmentation. Each of the segments focus on one particular hyperparameter and it’s major variations w.r.t other parameters possible.