unet_params

Typed parameters of the unet family; importable without Keras.

UNetBlockParams()

UNet block parameters

model_config = ConfigDict(frozen=True, extra='forbid')

filters: int = Field(..., description='# filters')

Number of filters

depth: int = Field(default=1, description='Layer depth')

Layer depth

ddepth: int | None = Field(default=None, description='Decoder depth')

Decoder depth

kernel: int | tuple[int, int] = Field(default=3, description='Kernel size')

Kernel size

pool: int | tuple[int, int] = Field(default=3, description='Pool size')

Pool size

strides: int | tuple[int, int] = Field(default=1, description='Stride size')

Stride size

skip: bool = Field(default=True, description='Add skip connection')

Add skip connection

seperable: bool = Field(default=False, description='Use seperable convs')

Use seperable convs

dropout: float | None = Field(default=None, description='Dropout rate')

Dropout rate

norm: Literal['batch', 'layer'] | None = Field(default='batch', description='Normalization type')

Normalization type

activation: Literal['relu', 'relu6', 'leaky_relu', 'elu', 'selu'] = Field(default='relu6', description='Activation')

Activation

dilation: int | tuple[int, int] | None = Field(default=None, description='Dilation factor')

Dilation factor

UNetParams()

UNet parameters

model_config = ConfigDict(frozen=True, extra='forbid')

family: Literal['unet'] = 'unet'

Model family

blocks: list[UNetBlockParams] = Field(default_factory=list, description='UNet blocks')

UNet blocks

include_top: bool = Field(default=True, description='Include top')

Include top

num_classes: int | None = Field(default=None, gt=0, description='Classes of the output layer; required with include_top')

Classes of the output layer; required with include_top

use_logits: bool = Field(default=True, description='Use logits')

Use logits

output_kernel_size: int | tuple[int, int] = Field(default=3, description='Output kernel size')

Output kernel size

output_kernel_stride: int | tuple[int, int] = Field(default=1, description='Output kernel stride')

Output kernel stride