Class SegResEncoder

monai/networks/nets/segresnet_ds.py:127–231

SegResEncoder based on the encoder structure in `3D MRI brain tumor segmentation using autoencoder regularization `_. Args: spatial_dims: spatial dimension of the input data. Defaults to 3. init_filters: number of output channels fo

class SegResEncoder(nn.Module):
    """
    SegResEncoder based on the encoder structure in `3D MRI brain tumor segmentation using autoencoder regularization
    <https://arxiv.org/pdf/1810.11654.pdf>`_.

    Args:
        spatial_dims: spatial dimension of the input data. Defaults to 3.
        init_filters: number of output channels for initial convolution layer. Defaults to 32.
        in_channels: number of input channels for the network. Defaults to 1.
        out_channels: number of output channels for the network. Defaults to 2.
        act: activation type and arguments. Defaults to ``RELU``.
        norm: feature normalization type and arguments. Defaults to ``BATCH``.
        blocks_down: number of downsample blocks in each layer. Defaults to ``[1,2,2,4]``.
        head_module: optional callable module to apply to the final features.
        anisotropic_scales: optional list of scale for each scale level.
    """

    def __init__(
        self,
        spatial_dims: int = 3,
        init_filters: int = 32,
        in_channels: int = 1,
        act: tuple | str = "relu",
        norm: tuple | str = "batch",
        blocks_down: tuple = (1, 2, 2, 4),
        head_module: nn.Module | None = None,
        anisotropic_scales: tuple | None = None,
    ):
        super().__init__()

        if spatial_dims not in (1, 2, 3):
            raise ValueError("`spatial_dims` can only be 1, 2 or 3.")

        # ensure normalization has affine trainable parameters (if not specified)
        norm = split_args(norm)
        if has_option(Norm[norm[0], spatial_dims], "affine"):
            norm[1].setdefault("affine", True)  # type: ignore

        # ensure activation is inplace (if not specified)
        act = split_args(act)
        if has_option(Act[act[0]], "inplace"):
            act[1].setdefault("inplace", True)  # type: ignore

        filters = init_filters  # base number of features

        kernel_size, padding, _ = aniso_kernel(anisotropic_scales[0]) if anisotropic_scales else (3, 1, 1)
        self.conv_init = Conv[Conv.CONV, spatial_dims](
            in_channels=in_channels,
            out_channels=filters,
            kernel_size=kernel_size,
            padding=padding,
            stride=1,
            bias=False,
        )
        self.layers = nn.ModuleList()

        for i in range(len(blocks_down)):
            level = nn.ModuleDict()