i UddlmZddlmZmZmZddlZddlmZddlm Z ddl m Z ddl mZmZmZddlmZdd lmZd ed ed eeeffd Z d%d ed ededeeeefdeeeefdeeeeeeffdedededed eeeeefffdZ d&dede dd efdZ d'd ed ededeeeefdeeeefde ddeeeeeeffdedededed eeeeefffdZ d(d ed ededeeeefdeeeefdeeeeeeffdededee dd eeeffdZ d)d ed ededeeeefdeeeefdeeeeeeffdededeeeeeeefeed ffdee dd efd!Z d&d"ede dd efd#Z d*d ed ededeeeefdeeeefde ddeeeeeeffdededeed fdee dd efd$Z y)+)Sequence)ListOptionalUnionN)Tensor) functional)Literal)_gaussian_kernel_2d_gaussian_kernel_3d_reflection_pad_3d)_check_same_shapereducepredstargetreturnc|j|jk7r|j|j}t||t|jdvr&t d|jd|jd||fS)zUpdate and returns variables required to compute Structural Similarity Index Measure. Args: preds: Predicted tensor target: Ground truth tensor )zMExpected `preds` and `target` to have BxCxHxW or BxCxDxHxW shape. Got preds: z and target: .)dtypetor lenshape ValueError)rrs w/Volumes/fast/ai/experiments/voice-extract-mac/.venv/lib/python3.12/site-packages/torchmetrics/functional/image/ssim.py_ssim_check_inputsrsz {{fll"5;;'eV$ 5;;v%  ;;-}V\\N! E   &=gaussian_kernelsigma kernel_size data_rangek1k2return_full_imagereturn_contrast_sensitivityc X |jdk(} t|ts| rd|gznd|gz}t|ts| rd|gznd|gz}t|t|jdz k7r-t dt|dt|jt|dvrt dt|t|t|jdz k7r-t dt|dt|jt|dvrt dt||r | r t dt d |Drt d |d t d |Drt d |d |Kt|j|jz |j|jz }nWt|trGtj||d|d}tj||d|d}|d|dz }t||zd} t||zd} |j} |jd}|j}|Dcgc]}t!d|zdzdzdz}}|r|ddz dz}|ddz dz}n|ddz dz}|ddz dz}| r9|ddz dz}t#||||}t#||||}|rYt%||||| }nIt'j(|||||fd}t'j(|||||fd}|rt+||||| }|sItj,|dg||| tj.tj0||| z }tj2||||z||z||zf}| rt'j4||nt'j6||}|j9|jd}|djd}|djd}|d|dz}tj|d|z d}tj|d|z d}|d|z }d|j;|z| z}||zj;|| z} d|z| z|z||z| z| zz }!| r|| z }"| r|"d|| || | f}"n|"d|| || f}"|!j=|!jddj?d|"j=|"jddj?dfS|r0|!j=|!jddj?d|!fS|!j=|!jddj?dScc}w)a%Compute Structural Similarity Index Measure. Args: preds: estimated image target: ground truth image gaussian_kernel: If true (default), a gaussian kernel is used, if false a uniform kernel is used sigma: Standard deviation of the gaussian kernel, anisotropic kernels are possible. Ignored if a uniform kernel is used kernel_size: the size of the uniform kernel, anisotropic kernels are possible. Ignored if a Gaussian kernel is used data_range: Range of the image. If ``None``, it is determined from the image (max - min) k1: Parameter of SSIM. k2: Parameter of SSIM. return_full_image: If true, the full ``ssim`` image is returned as a second argument. Mutually exclusive with ``return_contrast_sensitivity`` return_contrast_sensitivity: If true, the contrast term is returned as a second argument. The luminance term can be obtained with luminance=ssim/contrast Mutually exclusive with ``return_full_image`` rz`kernel_size` has dimension zD, but expected to be two less that target dimensionality, which is: )r)r(zMExpected `kernel_size` dimension to be 2 or 3. `kernel_size` dimensionality: zWArguments `return_full_image` and `return_contrast_sensitivity` are mutually exclusive.c3:K|]}|dzdk(xs|dkyw)r)rN).0xs r z_ssim_update..ks% 5A1q5A: a sz8Expected `kernel_size` to have odd positive number. Got rc3&K|] }|dk yw)rNr+)r,ys rr.z_ssim_update..ns !5a165sz.Expected `sigma` to have positive number. Got r)minmaxg @g?reflect)mode)rdevice)groupsg)r2r.) ndim isinstancerrrranyr3r2tupletorchclamppowr6sizerintr r Fpadr onesprodtensorcatconv3dconv2dsplitrreshapemean)#rrrr r!r"r#r$r%r&is_3dc1c2r6channelrsgauss_kernel_sizepad_hpad_wpad_dkernel input_listoutputs output_list mu_pred_sq mu_target_sqmu_pred_target sigma_pred_sqsigma_target_sqsigma_pred_targetupperlowerssim_idx_full_imagecontrast_sensitivitys# r _ssim_updaterd.s#@ JJ!OE k8 ,+0a;-'a;-6G eX &$UG !ug+ ;3v||,q00*3{+;*<=fll+, .   ;v%[\_`k\l[m n   5zS&***3{+;*<=fll+, .   5z[\_`k\l[m n  8rss 5 55ST_S``abcc !5 !!I%PQRSSuyy{2FJJL6::<4OP J & Ez!}*Q-HVAJqMJ]Z]2 R*_a B R*_a B \\FjjmG KKE=BCUS1Ws]+a/!3UC"1%)a/"1%)a/Q!#)Q!#) Q!#)"5%>#FE5%@ (2CUESYZFeeUE59 JvueU;)L (2CUESYZF Wa6+6eFSV[V`V` LLE& AW  E655=&6/5SY>Z[J>Cahhz6':R\^dmtIuG-- A/KQ##A&Jq>%%a(L ^k!n4NKK A ;EMkk+a.<"?SIO#A7 !$$U+ +b 0E _ , 0 0 7" +D+DQ+GLQQRTUWkWsWs & &q )2X $r( "**+>+D+DQ+GLQQRTUWjjj  & &':'@'@'CR H M Mb QQyDs6V' similarities reduction)elementwise_meansumnoneNct||S)aApply the specified reduction to pre-computed structural similarity. Args: similarities: per image similarities for a batch of images. reduction: a method to reduce metric score over individual batch scores - ``'elementwise_mean'``: takes the mean - ``'sum'``: takes the sum - ``'none'`` or ``None``: no reduction will be applied Returns: The reduced SSIM score r)rerfs r _ssim_computerks$ , **rc t||\}}t||||||||| | } t| tr| \} } t | || fS| } t | |S)aCompute Structural Similarity Index Measure. Args: preds: estimated image target: ground truth image gaussian_kernel: If true (default), a gaussian kernel is used, if false a uniform kernel is used sigma: Standard deviation of the gaussian kernel, anisotropic kernels are possible. Ignored if a uniform kernel is used kernel_size: the size of the uniform kernel, anisotropic kernels are possible. Ignored if a Gaussian kernel is used reduction: a method to reduce metric score over labels. - ``'elementwise_mean'``: takes the mean - ``'sum'``: takes the sum - ``'none'`` or ``None``: no reduction will be applied data_range: the range of the data. If None, it is determined from the data (max - min). If a tuple is provided then the range is calculated as the difference and input is clamped between the values. k1: Parameter of SSIM. k2: Parameter of SSIM. return_full_image: If true, the full ``ssim`` image is returned as a second argument. Mutually exclusive with ``return_contrast_sensitivity`` return_contrast_sensitivity: If true, the constant term is returned as a second argument. The luminance term can be obtained with luminance=ssim/contrast Mutually exclusive with ``return_full_image`` Return: Tensor with SSIM score Raises: TypeError: If ``preds`` and ``target`` don't have the same data type. ValueError: If ``preds`` and ``target`` don't have ``BxCxHxW shape``. ValueError: If the length of ``kernel_size`` or ``sigma`` is not ``2``. ValueError: If one of the elements of ``kernel_size`` is not an ``odd positive number``. ValueError: If one of the elements of ``sigma`` is not a ``positive number``. Example: >>> from torchmetrics.functional.image import structural_similarity_index_measure >>> preds = torch.rand([3, 3, 256, 256]) >>> target = preds * 0.75 >>> structural_similarity_index_measure(preds, target) tensor(0.9219) )rrdr:r<rk)rrrr r!rfr"r#r$r%r&similarity_pack similarityimages r#structural_similarity_index_measurerpsw~'uf5ME6"    # O/5)+ EZ3U:: J Y //r normalizerelusimplec t||||||||d \} } |dk(r*tj| } tj| } | | fS)NT)r&rs)rdr=rs) rrrr r!r"r#r$rqsimrcs r_get_normalized_sim_and_csrw'sb!-    $( !C Fjjo$zz*>? $ $$rbetas.c 6g} |jdk(} t|ts| rd|gznd|gz}t|ts| rd|gznd|gz}|jddt |zks"|jddt |zkr't dt |ddt |zdt d t |d z dz} |jd| z|d d z kr*t dt |d |d d |d d z | zd|jd| z|d d z kr*t dt |d |d d |d d z | zdtt |D]} t|||||||||  \}}| j|t |dk(r-tj|d}tj|d}ft |dk(r-tj|d}tj|d}t d| d<tj| }| dk(r|d zdz }tj||j j#dd }||z}tj$|d S)aCompute Multi-Scale Structural Similarity Index Measure. Adapted from: https://github.com/jorge-pessoa/pytorch-msssim/blob/master/pytorch_msssim/__init__.py. Args: preds: estimated image target: ground truth image gaussian_kernel: If true, a gaussian kernel is used, if false a uniform kernel is used sigma: Standard deviation of the gaussian kernel kernel_size: size of the gaussian kernel reduction: a method to reduce metric score over labels. - ``'elementwise_mean'``: takes the mean - ``'sum'``: takes the sum - ``'none'`` or ``None``: no reduction will be applied data_range: Range of the image. If ``None``, it is determined from the image (max - min) k1: Parameter of structural similarity index measure. k2: Parameter of structural similarity index measure. betas: Exponent parameters for individual similarities and contrastive sensitives returned by different image resolutions. normalize: When MultiScaleSSIM loss is used for training, it is desirable to use normalizes to improve the training stability. This `normalize` argument is out of scope of the original implementation [1], and it is adapted from https://github.com/jorge-pessoa/pytorch-msssim instead. Raises: ValueError: If the image height or width is smaller then ``2 ** len(betas)``. ValueError: If the image height is smaller than ``(kernel_size[0] - 1) * max(1, (len(betas) - 1)) ** 2``. ValueError: If the image width is smaller than ``(kernel_size[0] - 1) * max(1, (len(betas) - 1)) ** 2``. rr(r)r8z)For a given number of `betas` parameters zH, the image height and width dimensions must be larger than or equal to rr1rz and kernel size z', the image height must be larger than z&, the image width must be larger than )rq)r)r))r)r)r)z(length of kernel_size is neither 2 nor 3rt)r6)axis)r9r:rr@rrr3rangerwappendrB avg_pool2d avg_pool3dr=stackrFr6viewrE)rrrr r!r"r#r$rxrqmcs_listrM _betas_div_rvrc mcs_stack mcs_weighteds r_multiscale_ssim_updaterCsh H JJ!OE k8 ,+0a;-'a;-6G eX &$UG !ug+ zz|B!s5z/)UZZ\"-=SZ-O7E |D(()SZ'8 ;  QUa)Q.J zz|B:%Q!);;7E |CTU`abUcTde56A!nq6HJ5V4WWX Z   zz|B:%Q!);;7E |CTU`abUcTde45@^a5G:4U3VVW Y  3u: $> 6?E; BPR^g% ! ! ,- { q LL/E\\&&1F   "LL 2E\\&)4FGH HHRL H%IH]a' LLy'7'7 8 = =b! DEe#L ::l ++r mcs_per_imagect||S)aApply the specified reduction to pre-computed multi-scale structural similarity. Args: mcs_per_image: per image similarities for a batch of images. reduction: a method to reduce metric score over individual batch scores - ``'elementwise_mean'``: takes the mean - ``'sum'``: takes the sum - ``'none'`` or ``None``: no reduction will be applied Returns: The reduced multi-scale structural similarity r)rrfs r_multiscale_ssim_computers$ - ++rc t| ts tdt| trtd| Ds td| r| dvr tdt ||\}}t ||||||||| | } t | |S)ae Compute `MultiScaleSSIM`_, Multi-scale Structural Similarity Index Measure. This metric is a generalization of Structural Similarity Index Measure by incorporating image details at different resolution scores. Args: preds: Predictions from model of shape ``[N, C, H, W]`` target: Ground truth values of shape ``[N, C, H, W]`` gaussian_kernel: If true, a gaussian kernel is used, if false a uniform kernel is used sigma: Standard deviation of the gaussian kernel kernel_size: size of the gaussian kernel reduction: a method to reduce metric score over labels. - ``'elementwise_mean'``: takes the mean - ``'sum'``: takes the sum - ``'none'`` or ``None``: no reduction will be applied data_range: the range of the data. If None, it is determined from the data (max - min). If a tuple is provided then the range is calculated as the difference and input is clamped between the values. k1: Parameter of structural similarity index measure. k2: Parameter of structural similarity index measure. betas: Exponent parameters for individual similarities and contrastive sensitivities returned by different image resolutions. normalize: When MultiScaleSSIM loss is used for training, it is desirable to use normalizes to improve the training stability. This `normalize` argument is out of scope of the original implementation [1], and it is adapted from https://github.com/jorge-pessoa/pytorch-msssim instead. Return: Tensor with Multi-Scale SSIM score Raises: TypeError: If ``preds`` and ``target`` don't have the same data type. ValueError: If ``preds`` and ``target`` don't have ``BxCxHxW shape``. ValueError: If the length of ``kernel_size`` or ``sigma`` is not ``2``. ValueError: If one of the elements of ``kernel_size`` is not an ``odd positive number``. ValueError: If one of the elements of ``sigma`` is not a ``positive number``. Example: >>> from torch import rand >>> from torchmetrics.functional.image import multiscale_structural_similarity_index_measure >>> preds = rand([3, 3, 256, 256]) >>> target = preds * 0.75 >>> multiscale_structural_similarity_index_measure(preds, target, data_range=1.0) tensor(0.9628) References: [1] Multi-Scale Structural Similarity For Image Quality Assessment by Zhou Wang, Eero P. Simoncelli and Alan C. Bovik `MultiScaleSSIM`_ z3Argument `betas` is expected to be of a type tuple.c3<K|]}t|tyw)N)r:float)r,betas rr.zAmultiscale_structural_similarity_index_measure..s+VPUJtU,CPUsz5Argument `betas` is expected to be a tuple of floats.rrzNArgument `normalize` to be expected either `None` or one of 'relu' or 'simple')r:r<rallrrr) rrrr r!rfr"r#r$rxrqrs r.multiscale_structural_similarity_index_measurersJ eU #NOO%+VPU+V(VPQQY&88ijj&uf5ME6+ v{JBPUW`M $M9 ==r)T? N{Gz?Q?FF)rg) TrrrgNrrFF)TrrNrrN)TrrNrrgǺ?g48EG?ga4?g??g9EGr?N) TrrrgNrrrrs)!collections.abcrtypingrrrr=rtorch.nnrrBtyping_extensionsr #torchmetrics.functional.image.utilsr r r torchmetrics.utilities.checksr "torchmetrics.utilities.distributedrr<rboolrrArdrkrprwrrrr+rrrs/%(( $%ll;5ffvv~9N,!+.-/>B#(-LR LR LRLR ' ( LR sHSM)* LR ueE5L&99:; LR LR LRLR"&LR 65( ()LRbCU++>?+ +0!+.-/BT>B#(-R0 R0 R0R0 ' ( R0 sHSM)* R0 >? R0ueE5L&99:;R0 R0 R0R0"&R0 65( ()R0p!+.-/>B59% % %% ' ( % sHSM)* % ueE5L&99:; % % % 012% 66>%>!+.-/>BQ6:!f, f, f,f, ' ( f, sHSM)* f, ueE5L&99:; f, f, f, ueUE5895;LL Mf,  012!f," #f,VCU,,>?, ,0!+.-/BT>BG5;P> P> P>P> ' ( P> sHSM)* P> >? P>ueE5L&99:;P> P> P>  P> 012P> P>r