3j0SSKrSSKJr SSKrSSKJs Jr SSKJr SSKJ r SSK J r J r SSK Jr S\S\S \S \R 4S jrSS \S \R 4S jjrS\S\S \\\\44Sjr"SS\R(5rSS\R S\S\S\S\S\S \R 4Sjjr"SS\R(5rg)N)Tuple)nn)KORNIA_CHECK_SHAPE)get_gaussian_kernel2dspatial_gradient)pikdkykxreturncJX-U-nSnX4::ap[[S5(d 0[l[RnURU5nUc[R "U5nXeU'UR X0X5$[R "U5R X0X5$)z"Return neigh2channels conv kernel.i _eye_cache)hasattr_get_reshape_kernelrgettorcheyeview)r r r numel _MAX_CACHEDcacheress Q/home/wildlama/miniconda3/lib/python3.13/site-packages/kornia/feature/siftdesc.pyrrs2E K *L99-/  *#..ii ;))E"C%Lxx2**yy$$U77ksizec[U5S- nU[R"U5R5S-U- R5- n[R"X"5US-- nU$)zReturn a weighted pooling kernel for SIFT descriptor. Args: ksize: kernel_size. Returns: the pooling kernel with shape :math:`(ksize, ksize)`. @g?)floatrarangeabsger)rks_2xc2kernels rget_sift_pooling_kernelr&5s],$D %,,u%++-3d:??A AC YYs D!G ,F Mr patch_sizenum_spatial_binscS[XS-- 5-nX-nUS-nUSU--US- - S- U-S-nXQ:wa[SUSUS35eX#U4$)zReturn a tuple with SIFT parameters. Args: patch_size: the given patch size. num_spatial_bins: the ggiven number of spatial bins. Returns: ksize, stride, pad. rz Patch size zC is incompatible with requested number of spatial bins z{ for SIFT descriptor. Usually it happens when patch size is too small for num_spatial_bins specified)int ValueError)r'r(rstridepadout_sizes rget_sift_bin_ksize_stride_padr1EsS!';<==E0FzC!c')UQY7!;FJH#*& !"4 4  # rc ^\rSrSrSrS\4SjrSS\S\S\S\S \ SS 4 U4S jjjr S\ R4S jr S\ R4S jrS\ RS\ R4SjrSrU=r$)SIFTDescriptor]unn.Module which computes SIFT descriptors of given patches. Args: patch_size: Input patch size in pixels. num_ang_bins: Number of angular bins. num_spatial_bins: Number of spatial bins. clipval: clipping value to reduce single-bin dominance rootsift: if ``True``, RootSIFT (Arandjelović et. al, 2012) is computed. Returns: SIFT descriptor of the patches with shape. Shape: - Input: :math:`(B, 1, \text{num_spatial_bins}, \text{num_spatial_bins})` - Output: :math:`(B, \text{num_ang_bins * num_spatial_bins ** 2})` Example: >>> input = torch.rand(23, 1, 32, 32) >>> SIFT = SIFTDescriptor(32, 8, 4) >>> descs = SIFT(input) # 23x128 r c URRSURSURSURSUR SUR S3 $)N(num_ang_bins=, num_spatial_bins=z , patch_size= , rootsift= , clipval=)) __class____name__ num_ang_binsr(r'rootsiftclipvalselfs r__repr__SIFTDescriptor.__repr__usk~~&&'( --./ $ 5 567//*+ '||nA  ' rr'r=r(r>r?Nc >[T U]5 SUlX lX0lXPlX@lXlURn[U5[R"S5- n[Xf4Xw4S5Ul [X5uUlUlUl[#URS9R5n[$R&"SSUR)S5UR)S54URUR4UR UR 4SS9UlUR*R,R.R1UR3SSUR)S5UR)S555 g) N绽|=rTrr*rF) kernel_sizer.paddingbias)super__init__epsr=r(r?r>r'rmathsqrtrgkr1 bin_ksize bin_strider/r&rConv2dsizepkweightdatacopy_reshape) rAr'r=r(r>r?kssigmanwr;s rrKSIFTDescriptor.__init__s" ( 0  $//Ry499S>1'5.$G6ST^6q3$( $4>> : @ @ B)) RWWQZ0OOT__5XXtxx(   !!"**Q2771:rwwqz"JKrcJURRR5$N)rTrUdetachr@s rget_pooling_kernel!SIFTDescriptor.get_pooling_kernelsww~~$$&&rc6URR5$r^)rOr_r@s rget_weighting_kernel#SIFTDescriptor.get_weighting_kernelsww~~rinputc [USSURUR/5 URSnURR UR 5R UR 5Ul[US5nUSS2SS2S4nUSS2SS2S4n[R"XD-XU--UR-5n[R"XTUR-5S[--nX`RRU5RU5R UR 5-n[!UR"5U-S[-- n[R$"U5n X- n XR"-n U S-UR"-n SU - U-n X-n[R&"[)SUR"5Vs/sHSnURX:HR UR 5U -X:HR UR 5U--5PMU snS5nUR+US5n[,R."US S 9n[R0"US [!UR255n[,R."US S 9nUR4(a6[R"[,R."USS 9UR-5nU$s snf) NB1rdiffr*r?rpg)rr'shaperTtodtypedevicerrrNrLatan2rrO expand_astype_asrr=floorcatrangerF normalizeclampr?r>)rArerggradsgxgymagorio_bigbo0_big_wo1_big_bo0_bigbo1_bigwo0_bigwo1_bigiang_binss rforwardSIFTDescriptor.forwardsW53/@dooEV"XYQ''**U[[),,U\\: / 1a7^ 1a7^jj27*TXX56kk"488m,sRx7GG%%c*2237::3::FFd''(3.#(;;;u%#...Q;$"3"33>S(.99q$"3"34 4A))%++6@GLCTCTUZU`U`CadkCkkl4    ==B';;x1-;;xeDLL.AB;;x1- ==zz!++h!")r+T皙?)r< __module__ __qualname____firstlineno____doc__strrBr,boolrrKrTensorr`rcr__static_attributes__ __classcell__r;s@rr3r3]s. #  ! LLL L  L  L LLB'ELL' ell #U\\#ell##rr3rer=r>r?c([XX4U5"U5$)zXCompute the sift descriptor. See :class: `~kornia.feature.SIFTDescriptor` for details. )r3)rer'r=r(r>r?s r sift_describers *4DPW XY^ __rc^\rSrSrSrS\4SjrSS\S\S\S\S \ S \S \SS 4U4S jjjr S\ R4Sjr S\ RS\ R4SjrSrU=r$)DenseSIFTDescriptorudnn.Module, which computes SIFT descriptor densely over the image. Args: num_ang_bins: Number of angular bins. (8 is default) num_spatial_bins: Number of spatial bins per descriptor (4 is default). You might want to set odd number and relevant padding to keep feature map size spatial_bin_size: Size of a spatial bin in pixels (4 is default) clipval: clipping value to reduce single-bin dominance rootsift: (bool) if True, RootSIFT (Arandjelović et. al, 2012) is computed stride: default 1 padding: default 0 Returns: torch.Tensor: DenseSIFT descriptor of the image Shape: - Input: (B, 1, H, W) - Output: (B, num_ang_bins * num_spatial_bins ** 2, (H+padding)/stride, (W+padding)/stride) Examples:: >>> input = torch.rand(2, 1, 200, 300) >>> SIFT = DenseSIFTDescriptor() >>> descs = SIFT(input) # 2x128x194x294 r cURRSURSURSURSUR SUR SURS3$)Nr6r7z, spatial_bin_size=r8z , stride=r9r:)r;r<r=r(spatial_bin_sizer>r.r?r@s rrBDenseSIFTDescriptor.__repr__s{~~&&'( --./ $ 5 567 $ 5 567 'kk]#||nA  ' rr=r(rr>r?r.rHNc >[T U]5 SUlXlX lX0lXPlX@lX`lXpl [UR S9R5nURSURSSURS5URS555 [R "SSURS5URS54SSURS5S-URS5S-4S 9n U R"R$R'UR(5 Xl[-XU5R5n URS U 5 [R "UXS--X"4URUR4SURUR4S 9n U R"R$R'UR.5 XlUR(R35Ulg) NrErF_bin_pooling_kernel_weightr*r)r*r*Fr)rGr.rIrH_poolingconv_weight)rJrKrLr=r(rr?r>r.r/r&rregister_bufferrXrSrrRrUrVrWrbin_pooling_kernelrr PoolingConvr__pooling_kernel) rAr=r(rr>r?r.rHr[rPwrr;s rrKDenseSIFTDescriptor.__init__s ( 0 0   %4+@+@ A G G I 92::aBGGTUJXZX_X_`aXb;cdYY RWWQZ0WWQZ1_bggajAo6   !!&&,,T-L-LM"4 AQ R X X Z 2B7ii  Q. .)<KK-XXtxx(   %%d&>&>?& $>>EEGrcUR$r^)rr@s rr`&DenseSIFTDescriptor.get_pooling_kernel4s###rrec [U/SQ5 UR5up#pEURRUR5RUR 5UlUR RUR5RUR 5Ul[US5nUSS2SS2S4nUSS2SS2S4n[R"Xw-X--UR-5n [R"XUR-5S[--n [UR5U -S[-- n [R"U 5n X- n XR-nUS-UR-nSU - U -nX-n[R "[#SUR5Vs/sHUnURUU:HRUR5U-UU:HRUR5U--5PMW snS5nUR U5n[$R&"USSS9R)S[UR*55n[$R&"USSS9nUR,(a6[R"[$R&"USS 9UR-5nU$s snf) N)rgrhHWrirr*rrjr)dimrmrl)rrSrrorprqrrrrNrLrrrrr=rurvrwrxryclamp_r?r>)rAre_B_CH_W_Hr{r|r}r~rrrrrrrrrr out_no_normouts rrDenseSIFTDescriptor.forward8sO5"67**,"&"9"9"<">u||L / 1a7^ 1a7^jj27*TXX56kk"488m,sRx7d''(3.#(;;;u%#...Q;$"3"33>S(.99 q$"3"34  5A''\%%ekk2W<1 ?P?PQVQ\Q\?]`g?gg5    &&x0 kk+1299!U4<<=PQkk#1* ==**Q[[2TXX=>C  s3AK) rrrr?rLr=r(r/r>rr.)rr+r+Trr*r*)r<rrrrrrBr,rrrKrrr`rrrrs@rrrs4  #   ! !0H0H0H 0H  0H  0H0H0H 0H0Hd$ELL$#U\\#ell##rr)r)rMtypingrrtorch.nn.functionalr functionalrxkornia.core.checkrkornia.filtersrrkornia.geometry.conversionsrr,rrr&r1Moduler3rrrrrrrs$  0B*8C8S8c8ell8. 3   cSUSVX[]`S`Ma0lRYYlb ` << ` ` ` `  `  ` \\ ` "))r