UNetR3DDecoder#

pydantic model vision_architectures.nets.unetr_3d_decoder.UNetR3DDecoderConfig[source]#

Bases: CustomBaseModel

Show JSON schema

{
   "title": "UNetR3DDecoderConfig",
   "type": "object",
   "properties": {
      "num_outputs": {
         "description": "The number of output channels",
         "title": "Num Outputs",
         "type": "integer"
      },
      "conv_kernel_size": {
         "anyOf": [
            {
               "type": "integer"
            },
            {
               "maxItems": 3,
               "minItems": 3,
               "prefixItems": [
                  {
                     "type": "integer"
                  },
                  {
                     "type": "integer"
                  },
                  {
                     "type": "integer"
                  }
               ],
               "type": "array"
            }
         ],
         "description": "The kernel size of the convolution layers",
         "title": "Conv Kernel Size"
      },
      "final_layer_kernel_size": {
         "anyOf": [
            {
               "type": "integer"
            },
            {
               "maxItems": 3,
               "minItems": 3,
               "prefixItems": [
                  {
                     "type": "integer"
                  },
                  {
                     "type": "integer"
                  },
                  {
                     "type": "integer"
                  }
               ],
               "type": "array"
            }
         ],
         "description": "The kernel size of the final layer",
         "title": "Final Layer Kernel Size"
      }
   },
   "required": [
      "num_outputs",
      "conv_kernel_size",
      "final_layer_kernel_size"
   ]
}

Config:

arbitrary_types_allowed: bool = True
extra: str = ignore
validate_default: bool = True
validate_assignment: bool = True
validate_return: bool = True

Fields:

conv_kernel_size (int | tuple[int, int, int])
final_layer_kernel_size (int | tuple[int, int, int])
num_outputs (int)

Validators:

field num_outputs: int [Required]#

The number of output channels

Validated by:

field conv_kernel_size: int | tuple[int, int, int] [Required]#

The kernel size of the convolution layers

Validated by:

field final_layer_kernel_size: int | tuple[int, int, int] [Required]#

The kernel size of the final layer

Validated by:

pydantic model vision_architectures.nets.unetr_3d_decoder.UNetR3DStageConfig[source]#

Bases: CustomBaseModel

Show JSON schema

{
   "title": "UNetR3DStageConfig",
   "type": "object",
   "properties": {
      "in_dim": {
         "description": "The number of input channels",
         "title": "In Dim",
         "type": "integer"
      },
      "out_dim": {
         "description": "The number of output channels",
         "title": "Out Dim",
         "type": "integer"
      },
      "in_patch_size": {
         "description": "The patch size of the input",
         "maxItems": 3,
         "minItems": 3,
         "prefixItems": [
            {
               "type": "integer"
            },
            {
               "type": "integer"
            },
            {
               "type": "integer"
            }
         ],
         "title": "In Patch Size",
         "type": "array"
      },
      "out_patch_size": {
         "description": "The patch size of the output",
         "maxItems": 3,
         "minItems": 3,
         "prefixItems": [
            {
               "type": "integer"
            },
            {
               "type": "integer"
            },
            {
               "type": "integer"
            }
         ],
         "title": "Out Patch Size",
         "type": "array"
      }
   },
   "required": [
      "in_dim",
      "out_dim",
      "in_patch_size",
      "out_patch_size"
   ]
}

Config:

arbitrary_types_allowed: bool = True
extra: str = ignore
validate_default: bool = True
validate_assignment: bool = True
validate_return: bool = True

Fields:

in_dim (int)
in_patch_size (tuple[int, int, int])
out_dim (int)
out_patch_size (tuple[int, int, int])

Validators:

field in_dim: int [Required]#

The number of input channels

Validated by:

field out_dim: int [Required]#

The number of output channels

Validated by:

field in_patch_size: tuple[int, int, int] [Required]#

The patch size of the input

Validated by:

field out_patch_size: tuple[int, int, int] [Required]#

The patch size of the output

Validated by:

pydantic model vision_architectures.nets.unetr_3d_decoder.UNetR3DConfig[source]#

Bases: CustomBaseModel

Show JSON schema

{
   "title": "UNetR3DConfig",
   "type": "object",
   "properties": {
      "in_channels": {
         "description": "The number of input channels",
         "title": "In Channels",
         "type": "integer"
      },
      "decoder": {
         "$ref": "#/$defs/UNetR3DDecoderConfig",
         "description": "The decoder configuration"
      },
      "stages": {
         "description": "The stage configurations",
         "items": {
            "$ref": "#/$defs/UNetR3DStageConfig"
         },
         "title": "Stages",
         "type": "array"
      }
   },
   "$defs": {
      "UNetR3DDecoderConfig": {
         "properties": {
            "num_outputs": {
               "description": "The number of output channels",
               "title": "Num Outputs",
               "type": "integer"
            },
            "conv_kernel_size": {
               "anyOf": [
                  {
                     "type": "integer"
                  },
                  {
                     "maxItems": 3,
                     "minItems": 3,
                     "prefixItems": [
                        {
                           "type": "integer"
                        },
                        {
                           "type": "integer"
                        },
                        {
                           "type": "integer"
                        }
                     ],
                     "type": "array"
                  }
               ],
               "description": "The kernel size of the convolution layers",
               "title": "Conv Kernel Size"
            },
            "final_layer_kernel_size": {
               "anyOf": [
                  {
                     "type": "integer"
                  },
                  {
                     "maxItems": 3,
                     "minItems": 3,
                     "prefixItems": [
                        {
                           "type": "integer"
                        },
                        {
                           "type": "integer"
                        },
                        {
                           "type": "integer"
                        }
                     ],
                     "type": "array"
                  }
               ],
               "description": "The kernel size of the final layer",
               "title": "Final Layer Kernel Size"
            }
         },
         "required": [
            "num_outputs",
            "conv_kernel_size",
            "final_layer_kernel_size"
         ],
         "title": "UNetR3DDecoderConfig",
         "type": "object"
      },
      "UNetR3DStageConfig": {
         "properties": {
            "in_dim": {
               "description": "The number of input channels",
               "title": "In Dim",
               "type": "integer"
            },
            "out_dim": {
               "description": "The number of output channels",
               "title": "Out Dim",
               "type": "integer"
            },
            "in_patch_size": {
               "description": "The patch size of the input",
               "maxItems": 3,
               "minItems": 3,
               "prefixItems": [
                  {
                     "type": "integer"
                  },
                  {
                     "type": "integer"
                  },
                  {
                     "type": "integer"
                  }
               ],
               "title": "In Patch Size",
               "type": "array"
            },
            "out_patch_size": {
               "description": "The patch size of the output",
               "maxItems": 3,
               "minItems": 3,
               "prefixItems": [
                  {
                     "type": "integer"
                  },
                  {
                     "type": "integer"
                  },
                  {
                     "type": "integer"
                  }
               ],
               "title": "Out Patch Size",
               "type": "array"
            }
         },
         "required": [
            "in_dim",
            "out_dim",
            "in_patch_size",
            "out_patch_size"
         ],
         "title": "UNetR3DStageConfig",
         "type": "object"
      }
   },
   "required": [
      "in_channels",
      "decoder",
      "stages"
   ]
}

Config:

arbitrary_types_allowed: bool = True
extra: str = ignore
validate_default: bool = True
validate_assignment: bool = True
validate_return: bool = True

Fields:

decoder (vision_architectures.nets.unetr_3d_decoder.UNetR3DDecoderConfig)
in_channels (int)
stages (list[vision_architectures.nets.unetr_3d_decoder.UNetR3DStageConfig])

Validators:

validate » all fields

field in_channels: int [Required]#

The number of input channels

Validated by:

field decoder: UNetR3DDecoderConfig [Required]#

The decoder configuration

Validated by:

field stages: list[UNetR3DStageConfig] [Required]#

The stage configurations

Validated by:

validator validate » all fields[source]#: Base method for validating the model after creation.

class vision_architectures.nets.unetr_3d_decoder.UNetR3DDecoder(config={}, checkpointing_level=0, **kwargs)[source]#

Bases: Module, PyTorchModelHubMixin

UNetR3DDecoder made using multiple conv and deconv blocks.

__init__(config={}, checkpointing_level=0, **kwargs)[source]#

Initialize the UNetR3DDecoder.

Parameters:

config (UNetR3DConfig) – An instance of the Config class that contains all the configuration parameters. It can also be passed as a dictionary and the instance will be created automatically.
checkpointing_level (int) – The level of checkpointing to use for activation checkpointing. Refer to ActivationCheckpointing for more details.
**kwargs – Additional keyword arguments for configuration.

forward(embeddings, scan)[source]#

Process the multi-scale input embeddings and scan datapoint.

Parameters:

embeddings – Tensor of shape (B, C, Z, Y, X) or (B, Z, Y, X, C) representing the input features.
scan – Tensor of shape (B, C, Z, Y, X) or (B, Z, Y, X, C) representing the input features.

Return type:

Tensor

Returns:

Tensor of shape (B, C, Z, Y, X) or (B, Z, Y, X, C) representing the output features.

static soft_dice_loss_fn(prediction, target, reduction='mean', ignore_index=-100, smooth=1e-05)[source]#

Both prediction and target should be of the form (batch_size, num_classes, depth, width, height).

prediction: probability scores for each class target: should be binary masks.

static cross_entropy_loss_fn(prediction, target, reduction='mean', ignore_index=-100, smooth=1e-05)[source]#

Both prediction and target should be of the form (batch_size, num_classes, depth, width, height).

prediction: probability scores for each class target: should be binary masks.

static loss_fn(prediction, target, reduction='mean', weight_dsc=1.0, weight_ce=1.0, ignore_index=-100, smooth=1e-05, return_components=False)[source]#

Both prediction and target should be of the form (batch_size, num_classes, depth, width, height).

prediction: probability scores for each class target: should be binary masks.