h}.SSKJrJr SSKrSSKrSSKrSSKJrJr SSKJ r SSK J r SSK J r \RRS \S \\S \4S j5rS(S \S\S\S\S\4 Sjjr"SS5rS\\S \4SjrS\S\\S \4Sjr\R.R0S\\S\\\\4S\S\S \\\\44 Sj5r\R.R0S\\\4S\\S \\4Sj5r\R.R0S \\S\\S!\\S"\S#\\\S$\\S \4S%j5r"S&S'\R>5r g)))OptionalUnionN)nnTensor)box_area)_log_api_usage_once) roi_alignlevelsunmerged_resultsreturnc NUSnURURpC[R"UR S5UR S5UR S5UR S54X4S9n[ [ U55Hn[R"X:H5SRSSSS5nURUR S5XR S5XR S5XR S55nURSXqU5nM U$)Nrr rdtypedevice) rrtorchzerossizerangelenwhereviewexpandscatter)r r first_resultrrreslevelindexs O/var/www/auris/envauris/lib/python3.13/site-packages/torchvision/ops/poolers.py_onnx_merge_levelsr#s #A&L && (;(;6 ++ Q**1-|/@/@/C\EVEVWXEYZbg Cs+,- FO,Q/44RAqA JJqM  # ( ( +  # ( ( +  # ( ( +   kk!UU$;<. Jk_mink_maxcanonical_scalecanonical_levelepsc[XX#U5$N) LevelMapper)r%r&r'r(r)s r"initLevelMapperr-%s u_s KKr$c X\rSrSrSrSS\S\S\S\S\4 SjjrS \\ S \ 4S jr S r g )r,/zDetermine which FPN level each RoI in a set of RoIs should map to based on the heuristic in the FPN paper. Args: k_min (int) k_max (int) canonical_scale (int) canonical_level (int) eps (float) r%r&r'r(r)c@XlX lX0lX@lXPlgr+)r%r&s0lvl0r))selfr%r&r'r(r)s r"__init__LevelMapper.__init__;s  !# r$boxlistsrc R[R"[R"UVs/sHn[U5PM sn55n[R"UR [R "X0R- 5-[R"URURS9-5n[R"X@RURS9nUR[R5UR- R[R5$s snf)z$ Args: boxlists (list[BoxList]) r)minmax)rsqrtcatrfloorr2log2r1tensorr)rclampr%r&toint64)r3r6boxlists target_lvlss r"__call__LevelMapper.__call__Is JJuyy8!L8(7"38!LM Nkk$))ejjWW.E"E UYU]U]efelelHm"mn kk+::4::N u{{+djj8<) __name__ __module__ __qualname____firstlineno____doc__intfloatr4listrrF__static_attributes__r$r"r,r,/sa  #          Jf J& Jr$r,boxescT[R"USS9nURURp2[R"[ U5VVs/sH3upE[R "USS2SS24XC[R US9PM5 snnSS9n[R"Xa/SS9nU$s snnf)Nr)dimr )rlayoutr)rr<rr enumerate full_likestrided)rU concat_boxesrribidsroiss r"_convert_to_roi_formatraWs99U*L ''););E ))dmnsdtudt\`\]1bqb51%--PV Wdtu  C 99c(a 0D K vs :B$ feature original_sizec"URSSn/n[X!5HkupE[U5[U5- nS[[R"U5R 5R 55-nURU5 Mm US$)Nrr)shapeziprQrr?r>roundappend)rbrcrpossible_scaless1s2 approx_scalescales r" _infer_scalerobs == D#%Od*Ry59, U5<< 5::<BBDEEu%+ 1 r$features image_shapesc(U(d [S5eSnSnUH!n[USU5n[USU5nM# XE4nUVs/sHn[X5PM n n[R"[R "U S[R S95R5*n [R"[R "U S[R S95R5*n [[U 5[U 5UUS9n X4$s snf)Nzimages list should not be emptyrr r8rr'r() ValueErrorr:rorr>r?float32itemr-rP) rprqr'r(max_xmax_yrforiginal_input_shapefeatscaleslvl_minlvl_max map_levelss r" _setup_scalesrms :;; E EE!He$E!He$">CK L84l468F Lzz%,,vay FGLLNNGzz%,,vbzGHMMOOG  G  G '' J  MsDx featmap_namescl/nUR5Hup4X1;dM URU5 M U$r+)itemsri)rr x_filteredkvs r" _filter_inputrs5J     a  r$r output_sizesampling_ratior{mapperc UbUc [S5e[U5n[U5nUS:Xa[USUUUSUS9$U"U5n[U5n USRSn USR USR p[R"U U 4U-U U S9n /n[[X55Hununn[R"X:H5SnUUn[UUUUUS9n[R"5(a"URURU 55 MrURU R 5U U'M [R"5(a [!X5n U $)a Args: x_filtered (List[Tensor]): List of input tensors. boxes (List[Tensor[N, 4]]): boxes to be used to perform the pooling operation, in (x1, y1, x2, y2) format and in the image reference size, not the feature map reference. The coordinate must satisfy ``0 <= x1 < x2`` and ``0 <= y1 < y2``. output_size (Union[List[Tuple[int, int]], List[int]]): size of the output sampling_ratio (int): sampling ratio for ROIAlign scales (Optional[List[float]]): If None, scales will be automatically inferred. Default value is None. mapper (Optional[LevelMapper]): If none, mapper will be automatically inferred. Default value is None. Returns: result (Tensor) z$scales and mapper should not be Noner r)r spatial_scalerr)rtrrar rfrrrrrYrgr torchvision _is_tracingrirAr#)rrUrrr{r num_levelsr`r num_rois num_channelsrrresulttracing_resultsr per_level_featurern idx_in_levelrois_per_levelresult_idx_in_levels r"_multiscale_roi_alignrs,~?@@ZJ !% (DQ qM # ))   E]F4yHa=&&q)LqM''A)=)=6 [[       FO-6s:7N-O))!5{{6?3A6 l+'  #)    " " $ $  " "#6#9#9%#@ A$7#9#9&,,#GF< -.P0  #F< Mr$c ^\rSrSrSr\\\\\S.r SSS.S\\ S\ \ \ \ \\ 4S \ S \ S \ 4 U4S jjjrS \\ \4S\\S\\ \ \ 4S\4SjrS\ 4SjrSrU=r$)MultiScaleRoIAligna  Multi-scale RoIAlign pooling, which is useful for detection with or without FPN. It infers the scale of the pooling via the heuristics specified in eq. 1 of the `Feature Pyramid Network paper `_. They keyword-only parameters ``canonical_scale`` and ``canonical_level`` correspond respectively to ``224`` and ``k0=4`` in eq. 1, and have the following meaning: ``canonical_level`` is the target level of the pyramid from which to pool a region of interest with ``w x h = canonical_scale x canonical_scale``. Args: featmap_names (List[str]): the names of the feature maps that will be used for the pooling. output_size (List[Tuple[int, int]] or List[int]): output size for the pooled region sampling_ratio (int): sampling ratio for ROIAlign canonical_scale (int, optional): canonical_scale for LevelMapper canonical_level (int, optional): canonical_level for LevelMapper Examples:: >>> m = torchvision.ops.MultiScaleRoIAlign(['feat1', 'feat3'], 3, 2) >>> i = OrderedDict() >>> i['feat1'] = torch.rand(1, 5, 64, 64) >>> i['feat2'] = torch.rand(1, 5, 32, 32) # this feature won't be used in the pooling >>> i['feat3'] = torch.rand(1, 5, 16, 16) >>> # create some random bounding boxes >>> boxes = torch.rand(6, 4) * 256; boxes[:, 2:] += boxes[:, :2] >>> # original image size, before computing the feature maps >>> image_sizes = [(512, 512)] >>> output = m(i, [boxes], image_sizes) >>> print(output.shape) >>> torch.Size([6, 5, 3, 3]) )r{r~rIrJrsrrrr'r(c>[TU]5 [U5 [U[5(aX"4nXlX0l[U5UlSUl SUl X@l XPl gr+) superr4r isinstancerPrrtuplerr{r~r'r()r3rrrr'r( __class__s r"r4MultiScaleRoIAlign.__init__ s` D! k3 ' '&4K*, - ..r$rrUrqrc,[XR5nURb URc.[ XCUR UR 5uUlUl[UUURURURUR5$)a Args: x (OrderedDict[Tensor]): feature maps for each level. They are assumed to have all the same number of channels, but they can have different sizes. boxes (List[Tensor[N, 4]]): boxes to be used to perform the pooling operation, in (x1, y1, x2, y2) format and in the image reference size, not the feature map reference. The coordinate must satisfy ``0 <= x1 < x2`` and ``0 <= y1 < y2``. image_shapes (List[Tuple[height, width]]): the sizes of each image before they have been fed to a CNN to obtain feature maps. This allows us to infer the scale factor for each one of the levels to be pooled. Returns: result (Tensor) ) rrr{r~rr'r(rrr)r3rrUrqrs r"forwardMultiScaleRoIAlign.forward!s&#1&8&89 ;; $//"9+8$*>*>@T@T, (DK%         KK OO   r$cURRSURSURSURS3$)Nz(featmap_names=z, output_size=z, sampling_ratio=))rrKrrr)r3s r"__repr__MultiScaleRoIAlign.__repr__CsM~~&&'t7I7I6JK++,,=d>Q>Q=RRS U r$)r(r'rr~rrr{)rKrLrMrNrOrrRrQr,__annotations__strrrPrr4dictrrrrS __classcell__)rs@r"rrs!F"*$u+!6h{F[\O # /Cy/3c DI56/ /  ///*  V   F|  5c?+    D #  r$rrH)!typingrrrtorch.fxrrrtorchvision.ops.boxesrutilsr r jitunusedrRr#rPrQr-r,rarofxwraprrrrrrModulerrTr$r"rs" *' vf&, L L LL L  L%J%JP$v,6&cu6l*.uS#X*?RUhk 4; #$6T#v+&tCyT&\PV P <PcP P T%[ ! P [ ! P PPfa a r$