Method __init__

monai/networks/nets/efficientnet.py:567–620

Initialize EfficientNet-B0 to EfficientNet-B7 models as a backbone, the backbone can be used as an encoder for segmentation and objection models. Compared with the class `EfficientNetBN`, the only different place is the forward function. This class refers to `PyTorc

(
        self,
        model_name: str,
        pretrained: bool = True,
        progress: bool = True,
        spatial_dims: int = 2,
        in_channels: int = 3,
        num_classes: int = 1000,
        norm: str | tuple = ("batch", {"eps": 1e-3, "momentum": 0.01}),
        adv_prop: bool = False,
    )

class EfficientNetBNFeatures(EfficientNet):

    def __init__(
        self,
        model_name: str,
        pretrained: bool = True,
        progress: bool = True,
        spatial_dims: int = 2,
        in_channels: int = 3,
        num_classes: int = 1000,
        norm: str | tuple = ("batch", {"eps": 1e-3, "momentum": 0.01}),
        adv_prop: bool = False,
    ) -> None:
        """
        Initialize EfficientNet-B0 to EfficientNet-B7 models as a backbone, the backbone can
        be used as an encoder for segmentation and objection models.
        Compared with the class `EfficientNetBN`, the only different place is the forward function.

        This class refers to `PyTorch image models <https://github.com/rwightman/pytorch-image-models>`_.

        """
        blocks_args_str = [
            "r1_k3_s11_e1_i32_o16_se0.25",
            "r2_k3_s22_e6_i16_o24_se0.25",
            "r2_k5_s22_e6_i24_o40_se0.25",
            "r3_k3_s22_e6_i40_o80_se0.25",
            "r3_k5_s11_e6_i80_o112_se0.25",
            "r4_k5_s22_e6_i112_o192_se0.25",
            "r1_k3_s11_e6_i192_o320_se0.25",
        ]

        # check if model_name is valid model
        if model_name not in efficientnet_params:
            model_name_string = ", ".join(efficientnet_params.keys())
            raise ValueError(f"invalid model_name {model_name} found, must be one of {model_name_string} ")

        # get network parameters
        weight_coeff, depth_coeff, image_size, dropout_rate, dropconnect_rate = efficientnet_params[model_name]

        # create model and initialize random weights
        super().__init__(
            blocks_args_str=blocks_args_str,
            spatial_dims=spatial_dims,
            in_channels=in_channels,
            num_classes=num_classes,
            width_coefficient=weight_coeff,
            depth_coefficient=depth_coeff,
            dropout_rate=dropout_rate,
            image_size=image_size,
            drop_connect_rate=dropconnect_rate,
            norm=norm,
        )

        # only pretrained for when `spatial_dims` is 2
        if pretrained and (spatial_dims == 2):
            _load_state_dict(self, model_name, progress, adv_prop)

    def forward(self, inputs: torch.Tensor):
        """