Segmentation Model

UNet

class fusionlab.segmentation.UNet(cin, num_cls, base_dim=64, spatial_dims=2)

__init__(cin, num_cls, base_dim=64, spatial_dims=2)

Base Unet :param cin: input channels :type cin: int :param num_cls: number of classes :type num_cls: int :param base_dim: 1st stage dim of conv output :type base_dim: int

class fusionlab.segmentation.TFUNet(*args, **kwargs)

__init__(num_cls, base_dim=64)

Base Unet :param num_cls: number of classes :type num_cls: int :param base_dim: 1st stage dim of conv output :type base_dim: int

ResUNet

class fusionlab.segmentation.ResUNet(cin, num_cls, base_dim=64, spatial_dims=2)

class fusionlab.segmentation.TFResUNet(*args, **kwargs)

UNet++

class fusionlab.segmentation.UNet2plus(cin, num_cls, base_dim, spatial_dims=2)

class fusionlab.segmentation.TFUNet2plus(*args, **kwargs)

TransUNet

class fusionlab.segmentation.TransUNet(in_channels=3, img_size=224, num_classes=2, zero_head=False, decoder_channels=[256, 128, 64, 16], hidden_size=768, n_skip=3, skip_channels=[512, 256, 64, 16], patch_size=(16, 16))

forward(x)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

UNETR

class fusionlab.segmentation.UNETR(in_channels, out_channels, img_size, feature_size=16, hidden_size=768, mlp_dim=3072, num_heads=12, pos_embed='fc', dropout_rate=0.0, spatial_dims=2)

UNETR based on: “Hatamizadeh et al., UNETR: Transformers for 3D Medical Image Segmentation <https://arxiv.org/abs/2103.10504>”

source code: Project-MONAI/MONAI

__init__(in_channels, out_channels, img_size, feature_size=16, hidden_size=768, mlp_dim=3072, num_heads=12, pos_embed='fc', dropout_rate=0.0, spatial_dims=2)

Parameters:

• in_channels (int) – dimension of input channels.

• out_channels (int) – dimension of output channels.

• img_size (int) – dimension of input image.

• feature_size (int) – dimension of network feature size.

• hidden_size (int) – dimension of hidden layer.

• mlp_dim (int) – dimension of feedforward layer.

• num_heads (int) – number of attention heads.

• pos_embed (str) – position embedding layer type.

• norm_name – feature normalization type and arguments.

• conv_block – bool argument to determine if convolutional block is used.

• res_block – bool argument to determine if residual block is used.

• dropout_rate (float) – faction of the input units to drop in ViT.

• spatial_dims (int) – number of spatial dimensions.

forward(x_in)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

SegFormer

class fusionlab.segmentation.SegFormer(num_classes=6, mit_encoder_type='B0')

SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers<https://arxiv.org/abs/2105.15203>

source code: sithu31296/semantic-segmentation

Parameters:

• num_classes (int) – number of classes to segment

• mit_encoder_type (str) – type of MiT encoder, one of [‘B0’, ‘B1’, ‘B2’, ‘B3’, ‘B4’, ‘B5’]

forward(inputs)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

Return type:

Tensor

HuggingFace Segmentation Model

class fusionlab.segmentation.HFSegmentationModel(model, num_cls=None, loss_fct=CrossEntropyLoss())

Base Hugginface-pytoch model wrapper class of the segmentation model

forward(x, labels=None)

Defines the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.