0hddZddlZddlmZddlmZddlmZddlmZddlmcm Z ddl m Z ddl m Z dd lmZdd lmZdd lmZdd lmZdd lmZdZidddddddddddddddddd d!d"d#d$d%d&d'd(d)d*d+d,d-d.d/d0d1d2d3d4d5d6Zd7gd7d8ggd9d:gd:gd:d;gd;gd<Zd=d>d?d@dAd?d:dBd?dCd>d?gd=d>d?d@dAd?d:dDd?dCd>d?gd=d>d?d@dEd?d:dFd?dCd>d?gd=d>d?d@dGd?d:dFd?dCd>d?gd=dAd?d@dHd?d:dId?dCdAd?gd=dAd?d@dFd?d:dJd?dCdAd?gd=dAd?d@dKd?d:dLd?dCdAd?gd<ZdMdNdOdPdQdRZdSZddTZ ddUedVedWefdXZ ddYe!dZed[e!fd\Z" dd]e!d^e!d_e#dUedVedWedZed`efdaZ$ ddZedVedUedWed`ef dbZ%dcZ&dde!fdeZ'dfZ( ddge!dWedheeee!fdieeeffdjZ)edkdl ddmZ*edndo ddpZ+edqdr ddsZ,edtdu ddvZ-edwdx ddyZ.edzd{ dd|Z/ed}d~ ddZ0edddZ1edddZ2ejdje2_ejgde*_ejgde+_ejgde,_ejgde-_ejgde._ejgde/_ejgde0_y)zResNet-RS models for TF-Keras. Reference: - [Revisiting ResNets: Improved Training and Scaling Strategies]( https://arxiv.org/pdf/2103.07579.pdf) N)Callable)Dict)List)Union)backend)layers)imagenet_utils)training) data_utils) layer_utils) keras_exportzGhttps://storage.googleapis.com/tensorflow/keras-applications/resnet_rs/zresnet-rs-101-i160.h5 544b3434d00efc199d66e9058c7f3379zresnet-rs-101-i160_notop.h5 82d5b90c5ce9d710da639d6216d0f979zresnet-rs-101-i192.h5 eb285be29ab42cf4835ff20a5e3b5d23zresnet-rs-101-i192_notop.h5 f9a0f6b85faa9c3db2b6e233c4eebb5bzresnet-rs-152-i192.h5 8d72a301ed8a6f11a47c4ced4396e338zresnet-rs-152-i192_notop.h5 5fbf7ac2155cb4d5a6180ee9e3aa8704zresnet-rs-152-i224.h5 31a46a92ab21b84193d0d71dd8c3d03bzresnet-rs-152-i224_notop.h5 dc8b2cba2005552eafa3167f00dc2133zresnet-rs-152-i256.h5 ba6271b99bdeb4e7a9b15c05964ef4adzresnet-rs-152-i256_notop.h5 fa79794252dbe47c89130f65349d654azresnet-rs-200-i256.h5 a76930b741884e09ce90fa7450747d5fzresnet-rs-200-i256_notop.h5 bbdb3994718dfc0d1cd45d7eff3f3d9czresnet-rs-270-i256.h5 20d575825ba26176b03cb51012a367a8zresnet-rs-270-i256_notop.h5 2c42ecb22e35f3e23d2f70babce0a2aazresnet-rs-350-i256.h5 f4a039dc3c421321b7fc240494574a68zresnet-rs-350-i256_notop.h5 6e44b55025bbdff8f51692a023143d66zresnet-rs-350-i320.h5 7ccb858cc738305e8ceb3c0140bee393 ab0c1f9079d2f85a9facbd2c88aa6079 ae0eb9bed39e64fc8d7e0db4018dc7e8 fe6217c32be8305b1889657172b98884 69d9d925319f00a8bdd4af23c04e4102 90daa68cd26c95aa6c5d25451e095529)zresnet-rs-350-i320_notop.h5zresnet-rs-420-i320.h5zresnet-rs-420-i320_notop.h5zresnet-rs-50-i160.h5zresnet-rs-50-i160_notop.h5)r%r'i@)2e^@) input_filters num_repeatsi$5H,WVarianceScalingg@fan_outtruncated_normal)scalemode distribution) class_nameconfiga-Instantiates the {name} architecture. Reference: [Revisiting ResNets: Improved Training and Scaling Strategies]( https://arxiv.org/pdf/2103.07579.pdf) For image classification use cases, see [this page for detailed examples]( https://keras.io/api/applications/#usage-examples-for-image-classification-models). For transfer learning use cases, make sure to read the [guide to transfer learning & fine-tuning]( https://keras.io/guides/transfer_learning/). Note: each TF-Keras Application expects a specific kind of input preprocessing. For ResNetRs, by default input preprocessing is included as a part of the model (as a `Rescaling` layer), and thus `tf.keras.applications.resnet_rs.preprocess_input` is actually a pass-through function. In this use case, ResNetRS models expect their inputs to be float tensors of pixels with values in the [0-255] range. At the same time, preprocessing as a part of the model (i.e. `Rescaling` layer) can be disabled by setting `include_preprocessing` argument to False. With preprocessing disabled ResNetRS models expect their inputs to be float tensors of pixels with values in the [-1, 1] range. Args: depth: Depth of ResNet network. input_shape: optional shape tuple. It should have exactly 3 inputs channels, and width and height should be no smaller than 32. E.g. (200, 200, 3) would be one valid value. bn_momentum: Momentum parameter for Batch Normalization layers. bn_epsilon: Epsilon parameter for Batch Normalization layers. activation: activation function. se_ratio: Squeeze and Excitation layer ratio. dropout_rate: dropout rate before final classifier layer. drop_connect_rate: dropout rate at skip connections. include_top: whether to include the fully-connected layer at the top of the network. block_args: list of dicts, parameters to construct block modules. model_name: name of the model. pooling: optional pooling mode for feature extraction when `include_top` is `False`. - `None` means that the output of the model will be the 4D tensor output of the last convolutional layer. - `avg` means that global average pooling will be applied to the output of the last convolutional layer, and thus the output of the model will be a 2D tensor. - `max` means that global max pooling will be applied. weights: one of `None` (random initialization), `'imagenet'` (pre-training on ImageNet), or the path to the weights file to be loaded. Note: one model can have multiple imagenet variants depending on input shape it was trained with. For input_shape 224x224 pass `imagenet-i224` as argument. By default, highest input shape weights are downloaded. input_tensor: optional TF-Keras tensor (i.e. output of `layers.Input()`) to use as image input for the model. classes: optional number of classes to classify images into, only to be specified if `include_top` is True, and if no `weights` argument is specified. classifier_activation: A `str` or callable. The activation function to use on the "top" layer. Ignored unless `include_top=True`. Set `classifier_activation=None` to return the logits of the "top" layer. include_preprocessing: Boolean, whether to include the preprocessing layer (`Rescaling`) at the bottom of the network. Note: Input image is normalized by ImageNet mean and standard deviation. Defaults to `True`. Returns: A `keras.Model` instance. cVtjd}d|fd}|S)z Conv2D block with fixed padding.conv_c dkDr t|}tjdk(rdnddt|S)NsamevalidF)filters kernel_sizestridespaddinguse_biaskernel_initializername) fixed_paddingrConv2DCONV_KERNEL_INITIALIZER)inputsrMrNrSrOs ]/var/www/html/engine/venv/lib/python3.12/site-packages/tf_keras/src/applications/resnet_rs.pyapplyz!Conv2DFixedPadding..applysR Q;"6;7F v}}#%lF6   rget_uid)rMrNrOrScounterrYs```` rXConv2DFixedPaddingr^s. |//'*wi   LrZ bn_momentum bn_epsilon activationcVtjd}d|fd}|S)zResNet-D type STEM block.stem_ctjdk(rdnd}tddddz|}tj|dz |}tj d z |}tdddd z|}tj|d z |}tj dz |}tddddz|}tj|dz |}tj dz |}tddddz|}tj|dz |}tj dz |}|S)N channels_lastr0rJ  _stem_conv_1rMrNrOrS_stem_batch_norm_1axismomentumepsilonrS _stem_act_1rS _stem_conv_2_stem_batch_norm_2 _stem_act_2r/ _stem_conv_3_stem_batch_norm_3 _stem_act_3 _stem_conv_4_stem_batch_norm_4 _stem_act_4)rimage_data_formatr^rBatchNormalization Activation)rWbn_axisxrar`r_rSs rXrYzSTEM..applys002oE!1 Aqtn7L   F % % ,,    EF  jtm/C DQ G Aqtn7L   F % % ,,    EF  jtm/C DQ G Aqtn7L   F % % ,,    EF  jtm/C DQ G Aqtn7L   F % % ,,    EF  jtm/C DQ GrZr[)r_r`rarSr]rYs```` rXSTEMrs1 |//'*wi 2h LrZ in_filtersse_ratio expand_ratioctjdk(rdndtjd}d|fd}|S)zSqueeze and Excitation block.rer0rJse_c tjdz|}dk(r|jdddf}ndd|jdf}tj|dz|}t dt dzz}tj |ddgddgtddd d z |}tj dzzddgddgtddd d z |}tj||gdzS)N _se_squeezerprJ _se_reshaper4rKTrelu _se_reduce)rMrNrOrRrPrQrarSsigmoid _se_expand _se_excite) rGlobalAveragePooling2DshapeReshapemaxintrUrVmultiply) rWr~se_shapenum_reduced_filtersr}rrrSrs rXrYzSE..apply1s5 DF ) )tm/C DV L a< Q*H1aggbk*H ?FNN8$*> ? B!!Sa()B%CD FMM'AF6 $    FMMAF6  $   { 1DEErZ)rrzr\)rrrrSr]rYr}s```` @rXSEr(sN,,./AaqG |//%(WI"F"FH LrZrMrOuse_projectionsurvival_probabilityc jtjd} d| f d} | S)z7Bottleneck block variant for residual networks with BN.block_0_c* tjdk(rdnd}|} rdz} dk(r;tjddd dz |}t |dd d z |}nt |d d z |}tj | d z |}t dd dz |}tj | dz |}tj  dz|}t d dz |}tj | dz |}tj  dz|}t dzdd dz |}tj | dz |}d cxkrdkrnnt dz|} r!tj d dz|}tj||g}tj  dz|S)Nrer0rJr4rg)rgrgrK_projection_pooling) pool_sizerOrPrS_projection_convri_projection_batch_normrk_conv_1 batch_norm_1_act_1rp_conv_2 _batch_norm_2_act_2_conv_3 _batch_norm_3r_se)rrS)NrJrJrJ_drop) noise_shaperS _output_act) rrzrAveragePooling2Dr^r{r|rDropoutAdd)rWr}shortcut filters_outr~rar`r_rMrSrrOrrs rXrYzBottleneckBlock..applyhs002oE!1 !A+K!|622$"" 55    -' ! 22   -' !# 22   v00$"44   H AD9.applys O!#!!5 #   q+& A $!%%'%9gaSN*  A rZr[) rMrOr2rr`r_rarrSr]rYs ````````` rX BlockGrouprs6 |//.1gY': LrZc$|t|z|z S)z@Get survival probability based on block number and initial rate.)float init_rate block_num total_blockss rXget_survival_probabilityrs uY' ', 66rZ target_limitcbtj}||krtj|yy)z9Increase default recursion limit to create larger models.N)sysgetrecursionlimitsetrecursionlimit)r current_limits rXallow_bigger_recursionrs+))+M|# l+$rZch|dz }|dz}||z }tj||f||ff|}|S)zKPad the input along the spatial dimensions independently of input size.rJrg)rP)r ZeroPadding2D)rWrN pad_totalpad_begpad_end padded_inputss rXrTrTsVaI1nG'!GF((7#gw%78 M rZdepth block_argsclassifier_activationct|}| dk(rt|}| d|} |Dcgc]}d| }}| h|dvs=tjjj | st d|d|d| | |vr|r|dk7rt d |d |tj|d d tj|| }| tj|}n/tj| stj| |}n| }tjdk(rdnd}|}|rK||dz }tjd|}|dk(r"tjgdgd||}t!||||}| t"|} t%| D]P\}}t'||dzt)| dz}t+|d||dk(rdnd|d||||d|dzd |}R|rstj,d!"|}|dkDrtj.|d#"|}tj0|| tj2||d$%|}nC| d&k(rtj,d!"|}n!| d'k(rtj4d("|}| t7j8| }n|}t;j<||| "}| |vrc| j?d)d*}| d)|}|s|d+z }|d,}t@|z} tCjD|| d-tF|.}!|jI|!|S| |jI| |Scc}w)/a' Build Resnet-RS model, given provided parameters. Args: depth: Depth of ResNet network. input_shape: optional shape tuple. It should have exactly 3 inputs channels, and width and height should be no smaller than 32. E.g. (200, 200, 3) would be one valid value. bn_momentum: Momentum parameter for Batch Normalization layers. bn_epsilon: Epsilon parameter for Batch Normalization layers. activation: activation function. se_ratio: Squeeze and Excitation layer ratio. dropout_rate: dropout rate before final classifier layer. drop_connect_rate: dropout rate at skip connections. include_top: whether to include the fully-connected layer at the top of the network. block_args: list of dicts, parameters to construct block modules. model_name: name of the model. pooling: optional pooling mode for feature extraction when `include_top` is `False`. - `None` means that the output of the model will be the 4D tensor output of the last convolutional layer. - `avg` means that global average pooling will be applied to the output of the last convolutional layer, and thus the output of the model will be a 2D tensor. - `max` means that global max pooling will be applied. weights: one of `None` (random initialization), `'imagenet'` (pre-training on ImageNet), or the path to the weights file to be loaded. Note- one model can have multiple imagenet variants depending on input shape it was trained with. For input_shape 224x224 pass `imagenet-i224` as argument. By default, highest input shape weights are downloaded. input_tensor: optional TF-Keras tensor (i.e. output of `layers.Input()`) to use as image input for the model. classes: optional number of classes to classify images into, only to be specified if `include_top` is True, and if no `weights` argument is specified. classifier_activation: A `str` or callable. The activation function to use on the "top" layer. Ignored unless `include_top=True`. Set `classifier_activation=None` to return the logits of the "top" layer. include_preprocessing: Boolean, whether to include the preprocessing layer (`Rescaling`) at the bottom of the network. Note - Input image is normalized by ImageNet mean and standard deviation. Defaults to `True`. Returns: A `tf.keras.Model` instance. Raises: ValueError: in case of invalid argument for `weights`, or invalid input shape. ValueError: if `classifier_activation` is not `softmax` or `None` when using a pretrained top layer. imagenetz-iz imagenet-iNzThe `weights` argument should be either `None` (random initialization), `'imagenet'` (pre-training on ImageNet, with highest available input shape), or the path to the weights file to be loaded. For ResNetRSz- the following weight variants are available z% (default=highest). Received weights=z-If using `weights` as `'imagenet'` or any of zH with `include_top` as true, `classes` should be 1000. Received classes=r&rf) default_sizemin_size data_formatrequire_flattenweights)r)tensorrrer0rJgp?)rB)g ףp= ?gv/?gCl?)gg٪?g.?gQ?)meanvariancerl)r_r`rargrr1rr2rr) rMrarOr2rr_r`rrSavg_poolrp top_dropout predictions)rarSavgrmax_pool-r_notopz.h5models)fnameorigin cache_subdir file_hash)%DEPTH_TO_WEIGHT_VARIANTSrtfiogfileexists ValueErrorr obtain_input_shaperrzrInputis_keras_tensor Rescaling Normalizationr BLOCK_ARGS enumeraterlenrrrvalidate_activationDenseGlobalMaxPooling2Dr get_source_inputsr ModelsplitBASE_WEIGHTS_URLr get_file WEIGHT_HASHES load_weights)"r input_shaper_r`rar dropout_ratedrop_connect_rate include_topr model_namepoolingr input_tensorclassesrinclude_preprocessingavailable_weight_variantsmax_input_shaper~weights_allow_list img_inputr} num_channelsrargsrrWmodelweights_input_shape weights_namefilename download_url weights_paths" rXResNetRSrsT!9 ?*78IR014MNqJqc*NN.'..."%%++2D2DW2M !'"+,-!!(  +  $$D $%& 'y *  !33--/# KLL{3 &&|4 L LI$I,,./AaqGA"7Q;/ -F  9 -a 0 1 $$*7 A J:  A & Z( 47'!eZ1,  J)!q&Qa]+#!!5a!eWA&    ( :F ) )z :1 = ! @|-@CA**+@'J FLL  5M   e >--:>qAA  :))z:1=A..|< NN61: 6E$$%mmC04$Q':&;<  H $L"^3''(2 !**!#H-  <( L   7# LOs Mz'keras.applications.resnet_rs.ResNetRS50zkeras.applications.ResNetRS50c0td|dd||||||d| S)zBuild ResNet-RS50 model.r(?z resnet-rs-50 rrrrrrrrrrrr rrrrrrrrr s rX ResNetRS50rs5 !3!3  rZz(keras.applications.resnet_rs.ResNetRS101zkeras.applications.ResNetRS101c0td|dd||||||d| S)zBuild ResNet-RS101 model.r)rrz resnet-rs-101rrrs rX ResNetRS101r5 !3"3  rZz(keras.applications.resnet_rs.ResNetRS152zkeras.applications.ResNetRS152c0td|dd||||||d| S)zBuild ResNet-RS152 model.r*rrz resnet-rs-152rrrs rX ResNetRS152r"r rZz(keras.applications.resnet_rs.ResNetRS200zkeras.applications.ResNetRS200c0td|dd||||||d| S)zBuild ResNet-RS200 model.r+皙?rz resnet-rs-200rrrs rX ResNetRS200r%;r rZz(keras.applications.resnet_rs.ResNetRS270zkeras.applications.ResNetRS270cFtdtd|dd||||||d| S)zBuild ResNet-RS-270 model.ir,r$rz resnet-rs-270rrrrs rX ResNetRS270r(Ys=4 !3"3  rZz(keras.applications.resnet_rs.ResNetRS350zkeras.applications.ResNetRS350cFtdtd|dd||||||d| S)zBuild ResNet-RS350 model.ir-r$皙?z resnet-rs-350rr'rs rX ResNetRS350r+xs=4 !3"3  rZz(keras.applications.resnet_rs.ResNetRS420zkeras.applications.ResNetRS420cFtdtd|dd||||||d| S)zBuild ResNet-RS420 model.ir.r*r$z resnet-rs-420) rrrrrrrrrrrr r'rs rX ResNetRS420r-s=4 !3"3  rZz-keras.applications.resnet_rs.preprocess_inputc|S)aA placeholder method for backward compatibility. The preprocessing logic has been included in the ResnetRS model implementation. Users are no longer required to call this method to normalize the input data. This method does nothing and only kept as a placeholder to align the API surface between old and new version of model. Args: x: A floating point `numpy.array` or a `tf.Tensor`. data_format: Optional data format of the image tensor/array. `None` means the global setting `tf.keras.backend.image_data_format()` is used (unless you changed it, it uses "channels_last"). Defaults to `None`. Returns: Unchanged `numpy.array` or `tf.Tensor`. )r~rs rXpreprocess_inputr0s ( HrZz/keras.applications.resnet_rs.decode_predictionsc0tj||S)N)top)r decode_predictions)predsr2s rXr3r3s  , ,U <rN)rrJN)rr5rr皙?N)rr5rrr6N)Nrr5rrrg?TNz resnet-rsNrNrsoftmaxT)TrrNNNr7T))4__doc__rtypingrrrrtensorflow.compat.v2compatv2r tf_keras.srcrrtf_keras.src.applicationsr tf_keras.src.enginer tf_keras.src.utilsr r tensorflow.python.util.tf_exportr rrrrrVBASE_DOCSTRINGr^rstrrrrboolrrrrrTrrrr"r%r(r+r-r0r3formatr/rZrXrGs"  !! 4()*:N?!#E?"#E  ?  "#E ?"#E?"#E?"#E?"#E? "#E!"?#$$F?#E>"D- 4  s    s Q/a0a0a0  Q/a0b1a0  Q/a0b1a0  Q/b1b1a0  Q/b1b1a0  Q/b1b1a0  Q/b1b1a0  K+ Z$*IX. ????FJN--$-;>-h"% h h hh h  h  hh h^"% - -  -  -- -`7 ,, " '+   2;#P P  PT#s(^$P !h/!Pf-/N   #6.0P   #6.0P   #6.0P   #6.0P   #8.0P   #8.0P   #8=> ? ,?@=A=,>>FF#** *= $+++? $+++? $+++? $+++? $+++? $+++? rZ