h3SSKrSSKJr SSKJr SSKJrJrJrJ r SSK r SSK J r J r SSK Jr SSKJr SS KJr S S KJr S S KJrJrJr S S KJr S SKJrJr S SKJrJ r SSKJr! SSK"J#r# SSK$J%r% SSK&J'r'J(r( SS/r)S\*S\*S\*S\S\ RV4S\ RX4 Sjr-S\*S\*S\S\ RV4S\ RX4Sjr.S\ RV4S jr/"S!S"\ RV5r0"S#S$\(5r1"S%S&\(5r2"S'S(\ RV5r3S)\ \Rh\Rj4S*\*S\S\ RV44S+jr6"S,S\5r7\"5\"S-\7Rp4S.\ Rr4S/9SS0S\ RrSSS1.S2\\7S3\:S4\\*S5\\ S6\\*S\\S\ RV4S7\S\'4S8jj55r;g)9N) OrderedDict)partial)AnyCallableOptionalUnion)nnTensor)Conv2dNormActivation)ObjectDetection)_log_api_usage_once) mobilenet)register_modelWeights WeightsEnum)_COCO_CATEGORIES)_ovewrite_value_paramhandle_legacy_interface)mobilenet_v3_largeMobileNet_V3_Large_Weights)_utils)DefaultBoxGenerator)_validate_trainable_layers)SSDSSDScoringHead%SSDLite320_MobileNet_V3_Large_Weightsssdlite320_mobilenet_v3_large in_channels out_channels kernel_size norm_layer.returnc [R"[UUUUU[RS9[R"XS55$)N)r#groupsr$activation_layerr)r Sequentialr ReLU6Conv2d)r!r"r#r$s \/var/www/auris/envauris/lib/python3.13/site-packages/torchvision/models/detection/ssdlite.py_prediction_blockr-sC ==  #!XX   +Q/  c [RnUS-n[R"[XSX#S9[UUSSUUUS9[XASX#S95$)Nrr)r#r$r(r )r#strider'r$r()r r*r)r )r!r"r$ activationintermediate_channelss r, _extra_blockr30sfJ(A- == A*   ! !(!'   !Q: ! r.convc`UR5Hn[U[R5(dM$[RR R URSSS9 URcMf[RR RURS5 M g)NgQ?)meanstd) modules isinstancer r+torchinitnormal_weightbias constant_)r4layers r, _normal_initrBIsj eRYY ' ' HHMM ! !%,,Sd ! Czz% '' C8  r.c ^\rSrSrS\\S\\S\S\S\R44U4Sjjr S\\ S \ \ \ 44S jr S rU=r$) SSDLiteHeadQr! num_anchors num_classesr$.cf>[TU]5 [XX45Ul[ XU5UlgN)super__init__SSDLiteClassificationHeadclassification_headSSDLiteRegressionHeadregression_head)selfr!rFrGr$ __class__s r,rKSSDLiteHead.__init__Rs/ #<[Wb#o 4[zZr.xr%cHURU5URU5S.$)N)bbox_regression cls_logits)rOrM)rPrSs r,forwardSSDLiteHead.forwardYs(#33A62215  r.)rMrO)__name__ __module__ __qualname____firstlineno__listintrr ModulerKr dictstrrW__static_attributes__ __classcell__rQs@r,rDrDQsl[9[379[KN[\dehjljsjses\t[ f $sF{*;  r.rDc f^\rSrSrS\\S\\S\S\S\R44U4Sjjr Sr U=r $) rL`r!rFrGr$.c >[R"5n[X5H$upgUR[ XcU-SU55 M& [ U5 [ TU]XS5 g)Nr r ModuleListzipappendr-rBrJrK) rPr!rFrGr$rVchannelsanchorsrQs r,rK"SSDLiteClassificationHead.__init__asV]]_ !$[!> H   /:OQRT^_ `"?Z  1r. rYrZr[r\r]r^rr r_rKrbrcrds@r,rLrL`sG2923792KN2\dehjljsjses\t22r.rLcb^\rSrSrS\\S\\S\S\R44U4Sjjr Sr U=r $)rNkr!rFr$.c >[R"5n[X5H%upVUR[ USU-SU55 M' [ U5 [ TU]US5 g)Nr rh)rPr!rFr$bbox_regrlrmrQs r,rKSSDLiteRegressionHead.__init__lsS==?!$[!> H OO-hG Q S T"?X 1%r.rorprds@r,rNrNks=&DI&DI&S[\_acajaj\jSk&&r.rNc ^\rSrSrS S\R S\S\S\R 4S\S\4 U4Sjjjr S \ S \ \ \ 44S jr S rU=r$) SSDLiteFeatureExtractorMobileNettbackbonec4_posr$. width_mult min_depthc |>^^[TU]5 [U5 XR(a [ S5e[ R "[ R "/USUQXRSP76[ R "XRSS/XS-SQ765UlUU4Sjn[ R"[USRU"S5U5[U"S5U"S5U5[U"S5U"S5U5[U"S5U"S5U5/5n[U5 Xpl g) Nz0backbone[c4_pos].use_res_connect should be Falserrc4>[T[UT-55$rI)maxr^)dr}r|s r,;SSDLiteFeatureExtractorMobileNet.__init__..sc)SZ-@Ar.i)rJrKruse_res_connect ValueErrorr r)blockfeaturesrir3r"rBextra) rPrzr{r$r|r} get_depthrrQs `` r,rK)SSDLiteFeatureExtractorMobileNet.__init__us D!   + +OP P MM H8GV, Hh.>.D.DQ.G H MM(*004 Nx 7M N  B  Xb\66 # SYs^Ys^ZHYs^Ys^ZHYs^Ys^ZH    U r.rSr%c */nURHnU"U5nURU5 M URHnU"U5nURU5 M [[ U5VVs/sHupE[ U5U4PM snn5$s snnfrI)rrkrr enumeratera)rPrSoutputrivs r,rW(SSDLiteFeatureExtractorMobileNet.forwards]]EaA MM! #ZZEaA MM!  If4EF4EDASVQK4EFGGFs.B )rr)g?)rYrZr[r\r r_r^rfloatrKr r`rarWrbrcrds@r,rxrxts|   ))S"))^,     B H HDf$5 H Hr.rxrztrainable_layersc URnS/[U5VVs/sHup4[USS5(dMUPM snn-[U5S- /-n[U5nSUs=::aU::d O [ S5eUS:Xa [U5OXVU- nUSUH+nUR 5HnUR S5 M M- [XSU5$s snnf)Nr_is_cnFrzYtrainable_layers should be in the range [0, {num_stages}], instead got {trainable_layers})rrgetattrlenr parametersrequires_grad_rx) rzrr$rb stage_indices num_stages freeze_before parameters r,_mobilenet_extractorrs   HC8)<\)<8UZ@[1)<\\`cdl`mpq`q_rrM]#J  .J .tuu%5%:CM [kNk@lM n} %I  $ $U +(& ,HB6G TT]s CCc B\rSrSr\"S\S\SSSSS00S S S S .S 9r\rSr g)rzShttps://download.pytorch.org/models/ssdlite320_mobilenet_v3_large_coco-a79551df.pthi}4)rrz]https://github.com/pytorch/vision/tree/main/references/detection#ssdlite320-mobilenetv3-largez COCO-val2017box_mapgL5@g-?gt*@zSThese weights were produced by following a similar training recipe as on the paper.) num_params categoriesmin_sizerecipe_metrics_ops _file_size_docs)url transformsmetaroN) rYrZr[r\rr rCOCO_V1DEFAULTrbror.r,rrsG a"!*ut!  n  G$Gr. pretrainedpretrained_backbone)weightsweights_backboneT)rprogressrGrtrainable_backbone_layersr$rrrGrrkwargsc [RU5n[R"U5nSU;a[R"S5 Ub&Sn[ SU[ URS55nOUcSn[USL=(d USLUSS5nUSLnUc[[RSS S 9n[SX1XWS .UD6nUc [U5 [UUU5nS n [[!S5V s/sHn S S/PM sn SSS9n ["R$"X5n U R'5n [ U 5[ U R(5:wa-[+S[ U 5S[ U R(535eSSSS/SQ/SQS.n0UEUEn[-UU U U4S[/XX%50UD6nUbUR1UR3USS95 U$s sn f)aSSDlite model architecture with input size 320x320 and a MobileNetV3 Large backbone, as described at `Searching for MobileNetV3 `__ and `MobileNetV2: Inverted Residuals and Linear Bottlenecks `__. .. betastatus:: detection module See :func:`~torchvision.models.detection.ssd300_vgg16` for more details. Example: >>> model = torchvision.models.detection.ssdlite320_mobilenet_v3_large(weights=SSDLite320_MobileNet_V3_Large_Weights.DEFAULT) >>> model.eval() >>> x = [torch.rand(3, 320, 320), torch.rand(3, 500, 400)] >>> predictions = model(x) Args: weights (:class:`~torchvision.models.detection.SSDLite320_MobileNet_V3_Large_Weights`, optional): The pretrained weights to use. See :class:`~torchvision.models.detection.SSDLite320_MobileNet_V3_Large_Weights` below for more details, and possible values. By default, no pre-trained weights are used. progress (bool, optional): If True, displays a progress bar of the download to stderr. Default is True. num_classes (int, optional): number of output classes of the model (including the background). weights_backbone (:class:`~torchvision.models.MobileNet_V3_Large_Weights`, optional): The pretrained weights for the backbone. trainable_backbone_layers (int, optional): number of trainable (not frozen) layers starting from final block. Valid values are between 0 and 6, with 6 meaning all backbone layers are trainable. If ``None`` is passed (the default) this value is set to 6. norm_layer (callable, optional): Module specifying the normalization layer to use. **kwargs: parameters passed to the ``torchvision.models.detection.ssd.SSD`` base class. Please refer to the `source code `_ for more details about this class. .. autoclass:: torchvision.models.detection.SSDLite320_MobileNet_V3_Large_Weights :members: sizez?The size of the model is already fixed; ignoring the parameter.NrGr[gMbP?r6)epsmomentum)rrr$ reduced_tail)@rrr g?gffffff?) min_ratio max_ratioz4The length of the output channels from the backbone z? do not match the length of the anchor generator aspect ratios g?i,)?rr) score_thresh nms_threshdetections_per_imgtopk_candidates image_mean image_stdheadT)r check_hashro)rverifyrwarningswarnrrrrrr BatchNorm2drrBrrrange det_utilsretrieve_out_channelsnum_anchors_per_location aspect_ratiosrrrDload_state_dictget_state_dict)rrrGrrr$r reduce_tailrzr_anchor_generatorr"rFdefaultsmodels r,r r sBp4::7CG1889IJ  WX+M;GLLYeLfHgh   :t;/t;=VXY[\! #d*KR^^F ! hnHX#!H D*E!H+EHqQFH+EQT`de228BL";;=K <C 0 > >??B3|CTBUVUVYZjZxZxVyUz {  !&$ H)X((F     K L     E g44hSW4XY L?,Fs.G )rsD#11 ,2(77$CH!-6$,# $'69GOPSUWU^U^P^G_]]$c(3PRPYPY>BZ_a_l_l29ryy9  "))  22&N&-Hryy-H`UI))9+@+@@AUUbii(U.K, @HH I+-G-U-UV @D!%=W=e=e/359u ; <uu# u 9: u (} u#ryy.12uu u  ur.