This is the very first unet pipeline from scratch
the training image and labels are nifti files, mhd files could be transformed to nifti files first.
$ python unet_training.py
the models would be saved in ./runs_dict
$ python unet_evaluation.py
the output masks would be saved in ./tempdir