QMRITools`
QMRITools`

MakeUnet

MakeUnet[nChannels,nClasses,dep,dimIn]

Generates a UNET with nChannels as input and nClasses as output. he number of parameter of the first convolution layer can be set with dep. The data dimensions can be 2D or 3D and each of the dimensions should be 16, 32, 48, 64, 80, 96, 112, 128, 144, 160, 176, 192, 208, 224, 240 or 256.

Details

  • The following options can be given:
  • BlockType"ResNet"BlockType is an option for MakeUnet. It specifies which block are used to build the network. Values can be "UNET", "ResNet", "UResNet", "DenseNet" or "UDenseNet".
    DropoutRate0.2DropoutRate is an option for MakeUnet. It specifies how musch dropout is used after each block. It is a value between 0 and 1, default is .2.
    NetworkDepth5NetworkDepth is an option for MakeUnet. It specifief how deep the UNET will be.
    DownsampleScheduleAutomaticDownsampleSchedule is an option for MakeUnet. It defines how the data is downsampled for each of the deeper layers of the Unet. By default is is a factor two for each layer. A custum schedual for a 4 layer 3D Unet could be {{2,2,2},{1,2,2},{2,2,2},{1,2,2}}.
    InputFilters32InputFilters is an option for MakeUnet. It defines the amount of convolutional filters of the the first UNET block.
    ActivationType"GELU"ActivationType is an option for MakeUnet. It sepecifies which activation layer is used in the network. It can be "LeakyRELU" or any type allowed by a "name" definition in ElementwiseLayer.

Examples