o ) i4@sddlmZddlZddlmZddlZddlmZddlm Z m Z ddl m Z m Z ddlmZddgZGd ddejZeZdS) ) annotationsN)Sequence) Convolution ResidualUnit)ActNorm)SkipConnectionUNetUnetcsjeZdZdZdddejejdddfd0fdd Zd1d#d$Z d2d&d'Z d3d(d)Z d2d*d+Z d4d.d/Z ZS)5r a] Enhanced version of UNet which has residual units implemented with the ResidualUnit class. The residual part uses a convolution to change the input dimensions to match the output dimensions if this is necessary but will use nn.Identity if not. Refer to: https://link.springer.com/chapter/10.1007/978-3-030-12029-0_40. Each layer of the network has a encode and decode path with a skip connection between them. Data in the encode path is downsampled using strided convolutions (if `strides` is given values greater than 1) and in the decode path upsampled using strided transpose convolutions. These down or up sampling operations occur at the beginning of each block rather than afterwards as is typical in UNet implementations. To further explain this consider the first example network given below. This network has 3 layers with strides of 2 for each of the middle layers (the last layer is the bottom connection which does not down/up sample). Input data to this network is immediately reduced in the spatial dimensions by a factor of 2 by the first convolution of the residual unit defining the first layer of the encode part. The last layer of the decode part will upsample its input (data from the previous layer concatenated with data from the skip connection) in the first convolution. this ensures the final output of the network has the same shape as the input. Padding values for the convolutions are chosen to ensure output sizes are even divisors/multiples of the input sizes if the `strides` value for a layer is a factor of the input sizes. A typical case is to use `strides` values of 2 and inputs that are multiples of powers of 2. An input can thus be downsampled evenly however many times its dimensions can be divided by 2, so for the example network inputs would have to have dimensions that are multiples of 4. In the second example network given below the input to the bottom layer will have shape (1, 64, 15, 15) for an input of shape (1, 1, 240, 240) demonstrating the input being reduced in size spatially by 2**4. Args: spatial_dims: number of spatial dimensions. in_channels: number of input channels. out_channels: number of output channels. channels: sequence of channels. Top block first. The length of `channels` should be no less than 2. strides: sequence of convolution strides. The length of `stride` should equal to `len(channels) - 1`. kernel_size: convolution kernel size, the value(s) should be odd. If sequence, its length should equal to dimensions. Defaults to 3. up_kernel_size: upsampling convolution kernel size, the value(s) should be odd. If sequence, its length should equal to dimensions. Defaults to 3. num_res_units: number of residual units. Defaults to 0. act: activation type and arguments. Defaults to PReLU. norm: feature normalization type and arguments. Defaults to instance norm. dropout: dropout ratio. Defaults to no dropout. bias: whether to have a bias term in convolution blocks. Defaults to True. According to `Performance Tuning Guide `_, if a conv layer is directly followed by a batch norm layer, bias should be False. adn_ordering: a string representing the ordering of activation (A), normalization (N), and dropout (D). Defaults to "NDA". See also: :py:class:`monai.networks.blocks.ADN`. Examples:: from monai.networks.nets import UNet # 3 layer network with down/upsampling by a factor of 2 at each layer with 2-convolution residual units net = UNet( spatial_dims=2, in_channels=1, out_channels=1, channels=(4, 8, 16), strides=(2, 2), num_res_units=2 ) # 5 layer network with simple convolution/normalization/dropout/activation blocks defining the layers net=UNet( spatial_dims=2, in_channels=1, out_channels=1, channels=(4, 8, 16, 32, 64), strides=(2, 2, 2, 2), ) Note: The acceptable spatial size of input data depends on the parameters of the network, to set appropriate spatial size, please check the tutorial for more details: https://github.com/Project-MONAI/tutorials/blob/master/modules/UNet_input_size_constrains.ipynb. Typically, when using a stride of 2 in down / up sampling, the output dimensions are either half of the input when downsampling, or twice when upsampling. In this case with N numbers of layers in the network, the inputs must have spatial dimensions that are all multiples of 2^N. Usually, applying `resize`, `pad` or `crop` transforms can help adjust the spatial size of input data. rgTNDA spatial_dimsint in_channels out_channelschannels Sequence[int]strides kernel_sizeSequence[int] | intup_kernel_size num_res_unitsact tuple | strnormdropoutfloatbiasbool adn_orderingstrreturnNonecstt|dkrtdt|t|d}|dkr!td|dkr.td|dt|tr=t||kr=tdt|trLt||krLtd |_|_ |_ |_ |_ |_ |_|_| _| _| _| _| _dfdd ||j j d_dS)Nz2the length of `channels` should be no less than 2.rz len(channels) - 1`, the last z$ values of strides will not be used.z9the length of `kernel_size` should equal to `dimensions`.z._create_blockT) r%rr&rrrrrr'rr!r()super__init__r) ValueErrorwarningswarn isinstancer dimensionsrrrrrrrrrrrrmodel)r5r rrrrrrrrrrrrdelta __class__r3r7r9js6   z UNet.__init__ down_pathr(up_pathr0cCst|t||S)a Returns the block object defining a layer of the UNet structure including the implementation of the skip between encoding (down) and decoding (up) sides of the network. Args: down_path: encoding half of the layer up_path: decoding half of the layer subblock: block defining the next layer in the network. Returns: block for this layer: `nn.Sequential(down_path, SkipConnection(subblock), up_path)` )nn Sequentialr)r5rCrDr0r6r6r7r-s zUNet._get_connection_blockr'c Csj|jdkrt|j||||j|j|j|j|j|j|jd }|St |j||||j|j|j|j|j|jd }|S)a Returns the encoding (down) part of a layer of the network. This typically will downsample data at some point in its structure. Its output is used as input to the next layer down and is concatenated with output from the next layer to form the input for the decode (up) part of the layer. Args: in_channels: number of input channels. out_channels: number of output channels. strides: convolution stride. is_top: True if this is the top block. r)rrsubunitsrrrrr)rrrrrrr) rrr>rrrrrrr)r5rrrr'modr6r6r7r+s8   zUNet._get_down_layercCs|||ddS)z Returns the bottom or bottleneck layer at the bottom of the network linking encode to decode halves. Args: in_channels: number of input channels. out_channels: number of output channels. r$F)r+)r5rrr6r6r7r*szUNet._get_bottom_layercCst|j||||j|j|j|j|j|o|jdkd|jd }|jdkr>t |j||d|j d|j|j|j|j||jd }t ||}|S)a Returns the decoding (up) part of a layer of the network. This typically will upsample data at some point in its structure. Its output is used as input to the next layer up. Args: in_channels: number of input channels. out_channels: number of output channels. strides: convolution stride. is_top: True if this is the top block. rT) rrrrrr conv_only is_transposedrr$) rrrGrrrrlast_conv_onlyr) rr>rrrrrrrrrrErF)r5rrrr'convrur6r6r7r,s>    zUNet._get_up_layerx torch.TensorcCs||}|S)N)r?)r5rNr6r6r7forward(s z UNet.forward)r rrrrrrrrrrrrrrrrrrrrrrrrr r!r")rCr(rDr(r0r(r!r() rrrrrrr'rr!r()rrrrr!r()rNrOr!rO)__name__ __module__ __qualname____doc__rPRELUrINSTANCEr9r-r+r*r,rP __classcell__r6r6rAr7r s U N * /) __future__rr;collections.abcrtorchtorch.nnrE"monai.networks.blocks.convolutionsrrmonai.networks.layers.factoriesrr"monai.networks.layers.simplelayersr__all__Moduler r r6r6r6r7s