Skip to content

config

xvr.config

RegistrarArgs dataclass

RegistrarArgs(
    crop: int = 0,
    subtract_background: bool = False,
    linearize: bool = False,
    equalize: bool = False,
    reducefn: str = "max",
    pattern: str = "*.dcm",
    reverse_x_axis: bool = False,
    renderer: str = "trilinear",
    voxel_shift: float = 0.0,
    scales: str = "8",
    n_itrs: str = "500",
    parameterization: str = "euler_angles",
    convention: str = "ZXY",
    lr_rot: float = 0.01,
    lr_xyz: float = 1.0,
    patience: int = 10,
    threshold: float = 0.0001,
    max_n_plateaus: int = 3,
    init_only: bool = False,
    saveimg: bool = False,
    verbose: int = 1,
)

Default arguments for registration.

TrainerArgs dataclass

TrainerArgs(
    renderer: str = "trilinear",
    orientation: str = "AP",
    reverse_x_axis: bool = False,
    model_name: str = "resnet18",
    norm_layer: str = "groupnorm",
    pretrained: bool = False,
    parameterization: str = "quaternion_adjugate",
    convention: str = "ZXY",
    unit_conversion_factor: float = 1000.0,
    p_augmentation: float = 0.333,
    lr: float = 0.0002,
    weight_ncc: float = 1.0,
    weight_geo: float = 0.01,
    weight_dice: float = 1.0,
    weight_mvc: float = 0,
    batch_size: int = 116,
    n_total_itrs: int = 1000000,
    n_warmup_itrs: int = 1000,
    n_grad_accum_itrs: int = 4,
    n_save_every_itrs: int = 1000,
    disable_scheduler: bool = False,
    reuse_optimizer: bool = False,
    invert: bool = False,
    num_workers: int = 4,
    pin_memory: bool = False,
    project: str = "xvr",
)

Default arguments for training.