import os

from config.augm import train_transform, val_transform

from config.paths import train_images_folder, train_labels_folder, train_images, train_labels

from semseg.data_loader import SemSegConfig

class SemSegMRIConfig(SemSegConfig):

    train_images = [os.path.join(train_images_folder, train_image)

                    for train_image in train_images]

    train_labels = [os.path.join(train_labels_folder, train_label)

                    for train_label in train_labels]

    val_images = None

    val_labels = None

    do_normalize = True

    batch_size = 16

    num_workers = 0

    pad_ref = (48, 64, 48)

    lr = 0.01

    epochs = 100

    low_lr_epoch = epochs // 3

    val_epochs = epochs // 5

    cuda = True

    num_outs = 3

    do_crossval = True

    num_folders = 5

    num_channels = 8

    transform_train = train_transform

    transform_val = val_transform

    net = "vnet"

LEARNING_RATE_REDUCTION_FACTOR = 2