a d(@sXddlmZddlZddlmZGdddejjZGdddeZGdddeZ dS) )CallableNcseZdZUdZeed<eed<deeeddfd d Z e j d d d Z e j e j e j dddZ e j e j e j dddZZS)_HausdorffERLossBasea"Base class for binary Hausdorff loss based on morphological erosion. This is an Hausdorff Distance (HD) Loss that based on morphological erosion,which provided a differentiable approximation of Hausdorff distance as stated in :cite:`karimi2019reducing`. The code is refactored on top of `here `__. Args: alpha: controls the erosion rate in each iteration. k: the number of iterations of erosion. reduction: Specifies the reduction to apply to the output: 'none' | 'mean' | 'sum'. 'none': no reduction will be applied, 'mean': the weighted mean of the output is taken, 'sum': the output will be summed. Returns: Estimated Hausdorff Loss. convmax_pool@ meanN)alphak reductionreturncs0t||_||_||_|d|dS)Nkernel)super__init__r r r Zregister_buffer get_kernel)selfr r r  __class__`/var/www/html/stable-diffusion-webui/venv/lib/python3.9/site-packages/kornia/losses/hausdorff.pyrs  z_HausdorffERLossBase.__init__r cCstdS),Get kernel for image morphology convolution.N)NotImplementedError)rrrrr$sz_HausdorffERLossBase.get_kernelpredtargetr cCs ||d}tj|j|j|jd}tj||j|jd}tj||jtjd}t| ddd}t |j D]}|j |||dd} | d} d| | dk<| | } | |  } | | dk} | }|r| | | | }t|| || } || |d|j}| }qj|S)Ndevicedtype)Zweightpaddinggroupsg?r)torchZ as_tensorr rrZ zeros_likeZ ones_likeboolintsizeranger rrZsqueezeanywherer )rrrboundr Zerodedmaskr"r ZdilationZerosionZ erosion_maxZ erosion_minZ_to_normZto_normZ_erosion_to_fillrrrperform_erosion(s&    z$_HausdorffERLossBase.perform_erosioncsjddjddkr>ddkr>ddksXtdjdjddkrvtdtfd d tdD}jd kr| }n2jd kr| }njd krnt djd|S)anCompute Hausdorff loss. Args: pred: predicted tensor with a shape of :math:`(B, C, H, W)` or :math:`(B, C, D, H, W)`. Each channel is as binary as: 1 -> fg, 0 -> bg. target: target tensor with a shape of :math:`(B, 1, H, W)` or :math:`(B, C, D, H, W)`. Returns: Estimated Hausdorff Loss. rNrr!zTPrediction and target need to be of same size, and target should not be one-hot.Got z and .zInvalid target value.csZg|]R}dd||dft|ktjdjjdtjdjjdqS)Nr!rr)r-r$r*tensorrr).0irrrrr bs z0_HausdorffERLossBase.forward..rsumnonez reduction `z` has not been implemented yet.) shaper' ValueErrormaxitemr$stackr(r rr4r)rrroutrr2rforwardMs. >      z_HausdorffERLossBase.forward)rrr)__name__ __module__ __qualname____doc__r__annotations__floatr&strrr$Tensorrr-r< __classcell__rrrrrs %rcsPeZdZdZejZedZ ej dddZ ej ej ej dfdd Z Z S) HausdorffERLossa=Binary Hausdorff loss based on morphological erosion. Hausdorff Distance loss measures the maximum distance of a predicted segmentation boundary to the nearest ground-truth edge pixel. For two segmentation point sets X and Y , the one-sided HD from X to Y is defined as: .. math:: hd(X,Y) = \max_{x \in X} \min_{y \in Y}||x - y||_2 Furthermore, the bidirectional HD is: .. math:: HD(X,Y) = max(hd(X, Y), hd(Y, X)) This is an Hausdorff Distance (HD) Loss that based on morphological erosion, which provided a differentiable approximation of Hausdorff distance as stated in :cite:`karimi2019reducing`. The code is refactored on top of `here `__. Args: alpha: controls the erosion rate in each iteration. k: the number of iterations of erosion. reduction: Specifies the reduction to apply to the output: 'none' | 'mean' | 'sum'. 'none': no reduction will be applied, 'mean': the weighted mean of the output is taken, 'sum': the output will be summed. Examples: >>> hdloss = HausdorffERLoss() >>> input = torch.randn(5, 3, 20, 20) >>> target = (torch.rand(5, 1, 20, 20) * 2).long() >>> res = hdloss(input, target) r!rcCs.tgdgdgdgg}|d}|dS)rrr!rr!r!r!g?N)r$r/)rcrossr rrrrszHausdorffERLoss.get_kernelrcs|dkr td|d||dkrJ|dkrJ|jtjksttd|dd|d|d t ||S) a8Compute Hausdorff loss. Args: pred: predicted tensor with a shape of :math:`(B, C, H, W)`. Each channel is as binary as: 1 -> fg, 0 -> bg. target: target tensor with a shape of :math:`(B, 1, H, W)`. Returns: Estimated Hausdorff Loss. zOnly 2D images supported. Got r.r!rz+Expect long type target value in range (0, z).(z, )) dimr7r8r'minrr$longrr<rrrrrrr<s *$zHausdorffERLoss.forward)r=r>r?r@r$Zconv2drnnZAdaptiveMaxPool2drrDrr<rErrrrrF{s # rFcsPeZdZdZejZedZ ej dddZ ej ej ej dfdd Z Z S) HausdorffERLoss3DaLBinary 3D Hausdorff loss based on morphological erosion. Hausdorff Distance loss measures the maximum distance of a predicted segmentation boundary to the nearest ground-truth edge pixel. For two segmentation point sets X and Y , the one-sided HD from X to Y is defined as: .. math:: hd(X,Y) = \max_{x \in X} \min_{y \in Y}||x - y||_2 Furthermore, the bidirectional HD is: .. math:: HD(X,Y) = max(hd(X, Y), hd(Y, X)) This is a 3D Hausdorff Distance (HD) Loss that based on morphological erosion, which provided a differentiable approximation of Hausdorff distance as stated in :cite:`karimi2019reducing`. The code is refactored on top of `here `__. Args: alpha: controls the erosion rate in each iteration. k: the number of iterations of erosion. reduction: Specifies the reduction to apply to the output: 'none' | 'mean' | 'sum'. 'none': no reduction will be applied, 'mean': the weighted mean of the output is taken, 'sum': the output will be summed. Examples: >>> hdloss = HausdorffERLoss3D() >>> input = torch.randn(5, 3, 20, 20, 20) >>> target = (torch.rand(5, 1, 20, 20, 20) * 2).long() >>> res = hdloss(input, target) r!rcCs\tgdgdgdgg}tgdgdgdgg}tj|||gddd}|dS)rrGrH)rrrr!)rLg$I$I?N)r$r/r:)rrIr+r rrrrszHausdorffERLoss3D.get_kernelrcs.|dkr td|dt||S)aACompute 3D Hausdorff loss. Args: pred: predicted tensor with a shape of :math:`(B, C, D, H, W)`. Each channel is as binary as: 1 -> fg, 0 -> bg. target: target tensor with a shape of :math:`(B, 1, D, H, W)`. Returns: Estimated Hausdorff Loss. zOnly 3D images supported. Got r.)rLr7rr<rOrrrr<s zHausdorffERLoss3D.forward)r=r>r?r@r$Zconv3drrPZAdaptiveMaxPool3drrDrr<rErrrrrQs #  rQ) typingrr$Ztorch.nnrPZjitZ ScriptModulerrFrQrrrrs  tC