Ei?&SSKrSSKrSSKrSSKJr SSKJrJrJrJ r SSK r SSK r SSK Jr SSKJrJr SSKJrJr SSKJr SS KJrJr SS KJrJrJrJ r J!r!J"r" "S S \5r#"S S\5r$"SS\$5r%"SS\$5r&"SS\5r'g)N)Sequence)AnyCallableOptionalUnion)one_hot) tree_flattentree_unflatten) transforms tv_tensors) functional)_RandomApplyTransform Transform)_check_sequence_input_parse_labels_getterhas_anyis_pure_tensor query_chw query_sizec ^\rSrSrSr\R rS\\ \ 44U4Sjjr SS\ S\ \ S\ \ S\ S \4 U4S jjjrS \S \ S \ S\ S\ 4 U4SjjrS\\ S\\ \ 44SjrS \ S\\ \ 4S\ 4SjrSrU=r$) RandomErasingaRandomly select a rectangle region in the input image or video and erase its pixels. This transform does not support PIL Image. 'Random Erasing Data Augmentation' by Zhong et al. See https://arxiv.org/abs/1708.04896 Args: p (float, optional): probability that the random erasing operation will be performed. scale (tuple of float, optional): range of proportion of erased area against input image. ratio (tuple of float, optional): range of aspect ratio of erased area. value (number or tuple of numbers): erasing value. Default is 0. If a single int, it is used to erase all pixels. If a tuple of length 3, it is used to erase R, G, B channels respectively. If a str of 'random', erasing each pixel with random values. inplace (bool, optional): boolean to make this transform inplace. Default set to False. Returns: Erased input. Example: >>> from torchvision.transforms import v2 as transforms >>> >>> transform = transforms.Compose([ >>> transforms.RandomHorizontalFlip(), >>> transforms.PILToTensor(), >>> transforms.ConvertImageDtype(torch.float), >>> transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225)), >>> transforms.RandomErasing(), >>> ]) returncf>[[TU] 5URcSS9$URS9$)Nrandom)value)dictsuper _extract_params_for_v1_transformr)self __class__s _/home/mande/repo/quber/.venv/lib/python3.13/site-packages/torchvision/transforms/v2/_augment.pyr .RandomErasing._extract_params_for_v1_transform3s8 G 4 6"jj0(  6:jj  pscaleratiorinplacec>[TU]US9 [U[R[ [ [45(d [S5e[U[ 5(aUS:wa [S5e[U[5(d [S5e[U[5(d [S5eUSUS:d USUS:a[R"S 5 USS:d USS:a [S 5eX l X0l[U[[ 45(a[!U5/UlO`[U[ 5(aSUlOC[U[[ 45(a"UVs/sHn[!U5PM snUlOX@lXPl[&R("[&R*"UR55Ulgs snf) N)r&z=Argument value should be either a number or str or a sequencerz&If value is str, it should be 'random'zScale should be a sequencezRatio should be a sequencerrz,Scale and ratio should be of kind (min, max)zScale should be between 0 and 1)r__init__ isinstancenumbersNumberstrtuplelist TypeError ValueErrorrwarningswarnr'r(intfloatrr)torchlogtensor _log_ratio)r!r&r'r(rr)vr"s r#r+RandomErasing.__init__9sm 1%'..#ud!CDD[\ \ eS ! !ex&7EF F%**89 9%**89 9 !HuQx U1Xa%8 MMH I 8a<58a<>? ?  ec5\ * *,DJ s # #DJ e} - -,12Eq%(E2DJJ ))ELL$<= 3s/Gr inptargskwargsc,>[U[R[R[R45(aB[ R "[U5RS[U5RS35 [TU]("X/UQ70UD6$)Nz:() is currently passing through inputs of type tv_tensors.z(. This will likely change in the future.) r,r BoundingBoxes KeyPointsMaskr4r5type__name__r _call_kernel)r!r r>r?r@r"s r#rGRandomErasing._call_kernel\s} dZ55z7K7KZ__] ^ ^ MM:&&'("4j1122Z\ w#JFtFvFFr% flat_inputsc r[U5up#nURb*[UR5SU4;a[SUS35eX4-nURn[ S5GHnU[ R"S5RURSURS5R5-n[ R"[ R"S5RUSUS55R5n [[[R"X-555n [[[R"X- 555n X:aX:dGMURc4[ R"X*U /[ R S9R#5n O([ R$"UR5SS2SS4n [ R&"SX:- S-SS9R5n [ R&"SXK- S-SS9R5n O SSX4S4uppn [)XXU S 9$) Nrz@If value is a sequence, it should have either a single value or z (number of inpt channels) r)dtypersize)ijhwr<)rrlenr3r;ranger8emptyuniform_r'itemexpr6roundmathsqrtfloat32normal_r:randintr)r!rIimg_cimg_himg_warea log_ratio_ erase_area aspect_ratiorRrSr<rPrQs r# make_paramsRandomErasing.make_paramsds' 4e :: !c$**oaZ&GRSXRYYst }OO rA A 7 7 1 tzzRS} U Z Z \\J 99 A''aLaL df E$))J$=>?@AE$))J$=>?@AI!)zz!KK1 U]]CKKMLL,Qd]; aQT:??AA aQT:??AA +.q%4MA!a!,,r%paramscrUSb0UR"[RU40UDSUR0D6nU$)Nr<r))rGFeraser)r!r>rjs r# transformRandomErasing.transforms5 #; "$$QWWdSfSdllSD r%)r;r)r(r'r)?)g{Gz?gQ?)g333333?gffffff @gF)rF __module__ __qualname____firstlineno____doc__ _transformsr_v1_transform_clsrr/rr r7rboolr+rrGr1rhro__static_attributes__ __classcell__r"s@r#rrs<$11 $sCx. !-!+ !> !>!> !>  !>  !>!>FGxGsG3GRUGZ]G$-tCy$-T#s(^$-Lc4S>cr%rc^\rSrSrSSSS.S\S\\SS4U4S jjjrS rS \ RS \4S jr S\ RS\S\ R4Sjr Sr U=r$)_BaseMixUpCutMix?Ndefault)alpha num_classes labels_getterrrrc >[TU]5 [U5Ul[R R [R"U/5[R"U/55UlX l [U5Ul gN) rr+r7rr8 distributionsBetar:_distrr_labels_getter)r!rrrr"s r#r+_BaseMixUpCutMix.__init__s^ 5\ ((--ellE7.CU\\SXRYEZ[ &2=Ar%c ^ [U5S:aUOUSn[U5up#URU5n[U[R R [ R[ R[ R5(a![[U5RS35eURU5m [T [R 5(d[S[T 5S35eT R"S;a[ST R$S35eT R"S:XaSUR&bFT R$S UR&:wa)[S T R$S S UR&S 35eT R"S:XaUR&c [S 5eT T R$SS.UR)[+X$5VVs/sHupVU(dMUPM snn5EnSU[-U 4Sj[/U555'[+X$5VVs/sHupVU(aUR1XW5OUPM! nnn[3X5$s snnfs snnf)NrrzD() does not support PIL images, bounding boxes, keypoints and masks.z%The labels must be a tensor, but got z instead.)rzlabels should be index based with shape (batch_size,) or probability based with shape (batch_size, num_classes), but got a tensor of shape rzYWhen passing 2D labels, the number of elements in last dimension must match num_classes:  != z$. You can Leave num_classes to None.z=num_classes must be passed if the labels are index-based (1D))labels batch_sizeTc3:># UHupUTLdM Uv M g7fr).0idxr>rs r# +_BaseMixUpCutMix.forward..s !`7M)#QUY_Q_##7Ms  )rTr _needs_transform_listrPILImager rBrDrCr3rErFrr,r8Tensorndimshaperrhzipnext enumerateror ) r!inputsrIspecneeds_transform_listr>needs_transformrj flat_outputsrs @r#forward_BaseMixUpCutMix.forwards;v;?q (0 #99+F ; 1I1I:??\f\p\p q q:&&''kl $$V,&%,,//DT&\NR[\] ] ;;f $--3\\N)E  ;;!  0 0 <bAQUYUeUeAeT<<#$D)9)9(:;56  ;;!  0 0 8\] ] ,,q/ 585[o5[14_n5[o eiT!`y7M!``a,/{+Q +Q'-rc[U[R5(aSOSnURU:wa[ SUSURS35eUR SU:wa[ SUR SSUS 35eg) NzExpected a batched input with z dims, but got z dimensions instead.rzRThe batch size of the image or video does not match the batch size of the labels: r.)r,r Videorr3r)r!r>rexpected_num_dimss r#_check_image_or_video&_BaseMixUpCutMix._check_image_or_videos!+D*2B2B!C!CA 99) )01B0C?SWS\S\R]]qr  ::a=J &d::a=/j\4  'r%labellamc"URS:Xa[XRS9nURR(dUR 5nUR SS5RSU- 5RURU55$)Nr)rrr) rrrrLis_floating_pointr7rollmul_add_mul)r!rrs r# _mixup_label_BaseMixUpCutMix._mixup_labelsh ::?E/?/?@E{{,,KKMEzz!Q$$S3Y/44UYYs^DDr%)rrrr)rFrrrsrtr7rr6r+rr8rrrryrzr{s@r#r}r}s}),4_hBB8C=BmqBB-2^ %,, s E%,,EE%,,EEr%r}cZ\rSrSrSrS\\S\\\44Sjr S\S\\\4S\4Sjr S r g ) MixUpaqApply MixUp to the provided batch of images and labels. Paper: `mixup: Beyond Empirical Risk Minimization `_. .. note:: This transform is meant to be used on **batches** of samples, not individual images. See :ref:`sphx_glr_auto_examples_transforms_plot_cutmix_mixup.py` for detailed usage examples. The sample pairing is deterministic and done by matching consecutive samples in the batch, so the batch needs to be shuffled (this is an implementation detail, not a guaranteed convention.) In the input, the labels are expected to be a tensor of shape ``(batch_size,)``. They will be transformed into a tensor of shape ``(batch_size, num_classes)``. Args: alpha (float, optional): hyperparameter of the Beta distribution used for mixup. Default is 1. num_classes (int, optional): number of classes in the batch. Used for one-hot-encoding. Can be None only if the labels are already one-hot-encoded. labels_getter (callable or "default", optional): indicates how to identify the labels in the input. By default, this will pick the second parameter as the labels if it's a tensor. This covers the most common scenario where this transform is called as ``MixUp()(imgs_batch, labels_batch)``. It can also be a callable that takes the same input as the transform, and returns the labels. rIrcX[[URRS55S9$)Nrr)rr7rsample)r!rIs r#rhMixUp.make_paramss!djj//3455r%r>rjcUSnXSLaURXS9$[U[R[R45(d[ U5(aUR XSS9 URSS5RSU- 5RURU55n[U[R[R45(a[R"XAS 9nU$U$) Nrrrrrrrrlike) rr,r rrrrrrrrwrap)r!r>rjroutputs r#roMixUp.transformsUm (# #$$T$3 3 z//1A1AB C C~VZG[G[  & &t|8L & MYYq!_))#)499$((3-HF$!1!1:3C3C DEE#;MKr%rN rFrrrsrtrur1rrr/rhroryrr%r#rrsH46tCy6T#s(^6c4S>cr%rcZ\rSrSrSrS\\S\\\44Sjr S\S\\\4S\4Sjr S r g ) CutMixiaApply CutMix to the provided batch of images and labels. Paper: `CutMix: Regularization Strategy to Train Strong Classifiers with Localizable Features `_. .. note:: This transform is meant to be used on **batches** of samples, not individual images. See :ref:`sphx_glr_auto_examples_transforms_plot_cutmix_mixup.py` for detailed usage examples. The sample pairing is deterministic and done by matching consecutive samples in the batch, so the batch needs to be shuffled (this is an implementation detail, not a guaranteed convention.) In the input, the labels are expected to be a tensor of shape ``(batch_size,)``. They will be transformed into a tensor of shape ``(batch_size, num_classes)``. Args: alpha (float, optional): hyperparameter of the Beta distribution used for mixup. Default is 1. num_classes (int, optional): number of classes in the batch. Used for one-hot-encoding. Can be None only if the labels are already one-hot-encoded. labels_getter (callable or "default", optional): indicates how to identify the labels in the input. By default, this will pick the second parameter as the labels if it's a tensor. This covers the most common scenario where this transform is called as ``CutMix()(imgs_batch, labels_batch)``. It can also be a callable that takes the same input as the transform, and returns the labels. rIrcr[URRS55n[U5up4[R "USS9n[R "USS9nS[ R"SU- 5-n[Xt-5n[Xs-5n [[R"XX- SS95n [[R"Xi- SS95n [[R"XX-US95n [[R"Xi-US95n XX4n[SX- X- -XC-- - 5n[XS 9$) NrrMrNrqrr)min)max)box lam_adjusted) r7rrrr8r_r[r\r6clampr)r!rIrHWr_xr_yrr_w_halfr_h_halfx1y1x2y2rrs r#rhCutMix.make_params*sDJJ%%b)*+&mmAD)mmAD) $))C#I& &qu:qu: S^3 4 S^3 4 S^3 4 S^3 4rSBG#8AE#BBC 77r%r>rjcXSLaURXSS9$[U[R[R45(d[ U5(aUR XSS9 USup4pVURSS5nUR5nUS XF2X524US XF2X524'[U[R[R45(a[R"XS 9nU$U$) Nrrrrrrrr.r) rr,r rrrrrcloner) r!r>rjrrrrrolledrs r#roCutMix.transform@s (# #$$Tn/E$F F z//1A1AB C C~VZG[G[  & &t|8L & M#E]NBBYYq!_FZZ\F(.sBE25/@(AF3ru$ %$!1!1:3C3C DEE#;MKr%rNrrr%r#rrsH68tCy8T#s(^8,c4S>cr%rc^\rSrSrSrS\\\\44U4SjjrS\ \ S\ \ \ 44Sjr S\ S \ \ \ 4S\ 4S jrS rU=r$) JPEGiSaApply JPEG compression and decompression to the given images. If the input is a :class:`torch.Tensor`, it is expected to be of dtype uint8, on CPU, and have [..., 3 or 1, H, W] shape, where ... means an arbitrary number of leading dimensions. Args: quality (sequence or number): JPEG quality, from 1 to 100. Lower means more compression. If quality is a sequence like (min, max), it specifies the range of JPEG quality to randomly select from (inclusive of both ends). Returns: image with JPEG compression. qualitycn>[TU]5 [U[5(a$[U[5(a [ S5eX/nO [ USSS9 SUSs=::aUSs=::aS::a3O O0[US[5(a[US[5(d[SU<35eXlg) Nzquality can't be boolr)r) req_sizesrrdz7quality must be an integer from 1 to 100, got quality =) rr+r,r6rxr2rr3r)r!rr"s r#r+ JPEG.__init__cs  gs # #'4(( 788(G !'9 EWQZ471:44GAJPS9T9TYcdklmdnpsYtYtWg[YZ Z r%rIrc[R"URSURSS-S5R5n[ US9$)Nrrrr)r8r_rrXr)r!rIrs r#rhJPEG.make_paramsqs=-- Qa11DbINNPG$$r%r>rjcDUR[RXSS9$)Nrr)rGrljpegrns r#roJPEG.transformus!  i7H IIr%r)rFrrrsrtrurr6rr+r1rrr/rhroryrzr{s@r#rrSsm  c8C=&8 9 %tCy%T#s(^%JcJ4S>JcJJr%r)(r[r-r4collections.abcrtypingrrrr PIL.Imagerr8torch.nn.functionalrtorch.utils._pytreer r torchvisionr rvr torchvision.transforms.v2r rl _transformrr_utilsrrrrrrrr}rrrrr%r#rs $11 '<=58oo|)|~JEyJEZ- -`B BJ#J9#Jr%