model package

Subpackages

• Networks Architectures Package
  • Attention-UNet Architecture Module
  • Attention-UNet utils module
    • AttentionBlock
      • AttentionBlock.forward()
    • attention_up_block
      • attention_up_block.forward()
  • Layers module
    • BasicBlock
      • BasicBlock.forward()
    • Bottleneck
      • Bottleneck.forward()
    • ConvNormAct
      • ConvNormAct.forward()
    • DepthwiseSeparableConv
      • DepthwiseSeparableConv.forward()
  • MedFormer Architecture Module
    • MedFormer
      • MedFormer.forward()
  • MedFormer utils module
    • BasicLayer
      • BasicLayer.forward()
    • BidirectionAttention
      • BidirectionAttention.forward()
      • BidirectionAttention.rearrange1()
      • BidirectionAttention.rearrange2()
    • BidirectionAttentionBlock
      • BidirectionAttentionBlock.forward()
    • PatchMerging
      • PatchMerging.forward()
    • SemanticMapFusion
      • SemanticMapFusion.forward()
    • SemanticMapGeneration
      • SemanticMapGeneration.forward()
    • down_block
      • down_block.forward()
    • inconv
      • inconv.forward()
    • up_block
      • up_block.forward()
  • MedFormer utils-2 module
  • nnFormer Architecture Module
    • BasicLayer
      • BasicLayer.forward()
    • BasicLayer_up
      • BasicLayer_up.forward()
    • ContiguousGrad
      • ContiguousGrad.backward()
      • ContiguousGrad.forward()
    • Decoder
      • Decoder.forward()
    • Encoder
      • Encoder.forward()
    • Mlp
      • Mlp.forward()
    • PatchEmbed
      • PatchEmbed.forward()
    • PatchMerging
      • PatchMerging.forward()
    • Patch_Expanding
      • Patch_Expanding.forward()
    • SwinTransformerBlock
      • SwinTransformerBlock.forward()
    • SwinTransformerBlock_kv
      • SwinTransformerBlock_kv.forward()
    • WindowAttention
      • WindowAttention.forward()
    • WindowAttention_kv
      • WindowAttention_kv.forward()
    • final_patch_expanding
      • final_patch_expanding.forward()
    • nnFormer
      • nnFormer.forward()
    • project
      • project.forward()
    • window_partition()
    • window_reverse()
  • nnFormer utils module
    • InitWeights_He
    • InitWeights_XavierUniform
    • NeuralNetwork
      • NeuralNetwork.forward()
      • NeuralNetwork.get_device()
      • NeuralNetwork.set_device()
    • SegmentationNetwork
      • SegmentationNetwork.predict_2D()
      • SegmentationNetwork.predict_3D()
      • SegmentationNetwork.predict_3D_pseudo3D_2Dconv()
    • maybe_to_torch()
    • no_op
    • softmax_helper()
    • to_cuda()
  • SegFormer Architecture Module
  • SwinUNETR Architecture Module
    • BasicLayer
      • BasicLayer.forward()
    • PatchMerging
      • PatchMerging.forward()
    • PatchMergingV2
      • PatchMergingV2.forward()
    • SwinTransformer
      • SwinTransformer.forward()
      • SwinTransformer.proj_out()
    • SwinTransformerBlock
      • SwinTransformerBlock.forward()
      • SwinTransformerBlock.forward_part1()
      • SwinTransformerBlock.forward_part2()
      • SwinTransformerBlock.load_from()
    • SwinUNETR
      • SwinUNETR.forward()
      • SwinUNETR.load_from()
      • SwinUNETR.patch_size
    • WindowAttention
      • WindowAttention.forward()
    • window_partition()
    • window_reverse()
  • trans-layers module
    • Attention
      • Attention.forward()
      • Attention.rearrange1()
      • Attention.rearrange2()
    • LayerNorm
      • LayerNorm.forward()
    • Mlp
      • Mlp.forward()
    • TransformerBlock
      • TransformerBlock.forward()
  • UNet Architecture Module
    • UNet
      • UNet.forward()
  • UNet utils module
    • down_block
      • down_block.forward()
    • inconv
      • inconv.forward()
    • up_block
      • up_block.forward()
  • UNet++ Architecture Module
    • UNetPlusPlus
      • UNetPlusPlus.forward()
      • UNetPlusPlus.make_layer()
  • UNETR Architecture Module
    • UNETR
      • UNETR.forward()
      • UNETR.load_from()
      • UNETR.proj_feat()
  • utils module
    • get_act()
    • get_block()
    • get_norm()
  • VNet Architecture Module
    • ContBatchNorm3d
      • ContBatchNorm3d.forward()
    • DownTransition
      • DownTransition.forward()
    • ELUCons()
    • InputTransition
      • InputTransition.forward()
    • LUConv
      • LUConv.forward()
    • OutputTransition
      • OutputTransition.forward()
    • UpTransition
      • UpTransition.forward()
    • VNet
      • VNet.forward()
    • passthrough()
  • VTUNet Architecture Module
  • VTUNet utils module
  • Architecture selection module

model.utils module

model.utils.get_model(args, pretrain=False)

Function to get the model based on the arguments. Actually the models available, only 3 dimensions, are:

• UNETR

• UNet

• VNet

• AttentionUNet

• ResUNet

• MedFormer

• SegFormer

• UNetPlusPlus

• SwinUNETR

• SAM3D

• nnFormer: don’t work properly TODO

• VTUNet: don’t work properly TODO

• FCN_Net: don’t work properly TODO

TODO: Implement the other models, for example:

• DeepLabV3

• PSPNet

Parameters:

• args (argparse.Namespace) – Arguments from the command line.

• pretrain (bool, optional) – Set to true if you use a pretrained model. Defaults to False.

Raises:

• ValueError – No pretrain model available

• ValueError – Invalid dimension, should be ‘2d’ or ‘3d’

Returns:

The model object.

Return type:

Model

Module contents