2VhbddlZddlZddlmZddlmZddlmZddlmZddl m Z ddl m Z ddl mZd Zd d d d dddddZdddddddddddddddddddddddddddddddddddddddd ddddddd!d d"dddddddd"d#dddddgZd$d%d&d'd(d)Zd$d*d&d+d(d)Zd,Zgd-Z dUd.Z dVd/Zed0d1g dWd2Zed3d4g dXd5Zed6d7g dYd8Zed9d:g dZd;ZedZed?d@g d\dAZedBdCg d]dDZedEdFg d^dGZejAdHIe_!ejAdJIe_!ejAdKIe_!ejAdLIe_!ejAdMIe_!ejAdNIe_!ejAdOIe_!ejAdPIe_!edQd_dRZ"edSd`dTZ#ejFjBe#_!y)aN)backend)layers) keras_export)imagenet_utils) Functional)operation_utils) file_utilsz2https://storage.googleapis.com/keras-applications/) 902e53a9f72be733fc0bcb005b3ebbac 50bc09e76180e00e4465e1a485ddc09d) 1d254153d4ab51201f1646940f018540 74c4e6b3e1f6a1eea24c589628592432) b15cce36ff4dcbd00b6dd88e7857a6ad 111f8e2ac8aa800a7a99e3239f7bfb39) ffd1fdc53d0ce67064dc6a9c7960ede0 af6d107764bb5b1abb91932881670226) 18c95ad55216b8f92d7e70b3a046e2fc ebc24e6d6c33eaebbd558eafbeedf1ba) ace28f2a6363774853a83a0b21b9421a 38879255a25d3c92d5e44e04ae6cec6f) 165f6e37dce68623721b423839de8be5 9ecce42647a20130c1f39a5d4cb75743) 8c03f828fec3ef71311cd463b6759d99 cbcfe4450ddf6f3ad90b1b398090fe4a)b0b1b2b3b4b5b6b7 Tg?) kernel_sizerepeats filters_in filters_out expand_ratioid_skipstridesse_ratio(Ppi@VarianceScaling@fan_outtruncated_normal)scalemode distribution) class_nameconfiggUUUUUU?uniforma Instantiates the {name} architecture. Reference: - [EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks]( https://arxiv.org/abs/1905.11946) (ICML 2019) This function returns a Keras image classification model, optionally loaded with weights pre-trained on ImageNet. 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 Keras Application expects a specific kind of input preprocessing. For EfficientNet, input preprocessing is included as part of the model (as a `Rescaling` layer), and thus `keras.applications.efficientnet.preprocess_input` is actually a pass-through function. EfficientNet models expect their inputs to be float tensors of pixels with values in the `[0-255]` range. Args: include_top: Whether to include the fully-connected layer at the top of the network. Defaults to `True`. weights: One of `None` (random initialization), `"imagenet"` (pre-training on ImageNet), or the path to the weights file to be loaded. Defaults to `"imagenet"`. input_tensor: Optional Keras tensor (i.e. output of `layers.Input()`) to use as image input for the model. input_shape: Optional shape tuple, only to be specified if `include_top` is False. It should have exactly 3 inputs channels. pooling: Optional pooling mode for feature extraction when `include_top` is `False`. Defaults to `None`. - `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. 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. 1000 is how many ImageNet classes there are. Defaults to `1000`. 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. Defaults to `'softmax'`. When loading pretrained weights, `classifier_activation` can only be `None` or `"softmax"`. name: The name of the model (string). Returns: A model instance. )gZd;O?gy&1?g?c $ !|dk(rt}| dvs tj| s td| dk(r| r|dk7r tdt j | |dt j| | } | tj| }n/t j| stj| | }n| }t jd k(rd nd }|ffd }fd!|}tjd|}tj||}| dk(rCtjtDcgc]}dtj|z c}|}tj t j"|d d|}tj$|dd dddt&d|}tj(|d|}tj*|d|}t-j.|}d}t1t3!fd|D}t5|D]\}}|d dkDsJ||d!|d!<||d"|d"<t7!|j9d D]D}|dkDr d |d#<|d"|d!<t;||||z|z fd$d%|d zt=|d&zd'i|}|d z }Ftj$|d(d d)dt&d*+|}tj(|d,|}tj*|d-|}| rxtj>d.|}|dkDrtj@|d/|}t jB|| tjD||tFd01|}nC| d2k(rtj>d.|}n!| d3k(rtjHd4|}| tKjL| }n|}tO|||}| dk(rW| rd5}tP|d}nd6}tP|d }||z}tjR|tT|zd7|8} |jW| |S| |jW| |Scc}w)9aJInstantiates the EfficientNet architecture. Args: width_coefficient: float, scaling coefficient for network width. depth_coefficient: float, scaling coefficient for network depth. default_size: integer, default input image size. dropout_rate: float, dropout rate before final classifier layer. drop_connect_rate: float, dropout rate at skip connections. depth_divisor: integer, a unit of network width. activation: activation function. blocks_args: list of dicts, parameters to construct block modules. name: string, model name. include_top: whether to include the fully-connected layer at the top of the network. weights: one of `None` (random initialization), 'imagenet' (pre-training on ImageNet), or the path to the weights file to be loaded. input_tensor: optional Keras tensor (i.e. output of `layers.Input()`) to use as image input for the model. input_shape: optional shape tuple, only to be specified if `include_top` is False. It should have exactly 3 inputs channels. 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. 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. Returns: A model instance. default>NimagenetzThe `weights` argument should be either `None` (random initialization), `imagenet` (pre-training on ImageNet), or the path to the weights file to be loaded.rCzRIf using `weights="imagenet"` with `include_top` as true, `classes` should be 1000r$) default_sizemin_size data_formatrequire_flattenweights)shape)tensorrJ channels_lastr"r#c|z}t|t||dz z|z|z}|d|zkr||z }t|S)z2Round number of filters based on depth multiplier.r.g?)maxint)filtersdivisor new_filterswidth_coefficients S/home/dcms/DCMS/lib/python3.12/site-packages/keras/src/applications/efficientnet.py round_filtersz#EfficientNet..round_filters4sX$$ S7Q;./7:WD  w & 7 "K;cFttj|zS)z2Round number of repeats based on depth multiplier.)rOmathceil)r'depth_coefficients rT round_repeatsz#EfficientNet..round_repeats?s499.89::rVgp?)axis? stem_conv_padpaddingnamer.validF stem_conv)r,r`use_biaskernel_initializerrastem_bnr\rastem_activationrarc34K|]}|dyw)r'N).0argsr[s rT zEfficientNet..gsN$}T)_5Nsr'r(r)r,rablocka_isametop_convr`rdreratop_bntop_activationavg_pool top_dropout predictions) activationreraavgrNmax_poolz.h5z _notop.h5models) cache_subdir file_hash),DEFAULT_BLOCKS_ARGSr exists ValueErrorrobtain_input_shaperimage_data_formatrInputis_keras_tensor Rescaling NormalizationIMAGENET_STDDEV_RGBrXsqrt ZeroPadding2D correct_padConv2DCONV_KERNEL_INITIALIZERBatchNormalization Activationcopydeepcopyfloatsum enumeraterangepoprochrGlobalAveragePooling2DDropoutvalidate_activationDenseDENSE_KERNEL_INITIALIZERGlobalMaxPooling2Drget_source_inputsrWEIGHTS_HASHESget_fileBASE_WEIGHTS_PATH load_weights)"rSrZrE dropout_ratedrop_connect_rate depth_divisorrz blocks_argsra include_toprI input_tensor input_shapepoolingclassesclassifier_activation weights_name img_inputbn_axisrUxstddevbblocksirmjinputsmodel file_suffixr file_name weights_pathr[s"`` @rT EfficientNetrs~i) ) )Z->->w-G <  *D 1  !33!--/# KLL{3 &&|4 L LI$I,,./AaqG'4  ; A%%a(A*'*1-A* F  3F GS499V$ $ G   **1a0  A  b 2   A @!!wY?BA=*+<=a@A-- ,K A 3N+NN OF[)4I"""*4 +=>\+D,?@]}TXXi%89: A1u"#Y%)-%8\"!A%.QUGCBK=2   A FA  ,  d 2    A ?!!wX>qAA<*+;--:>qAA  :))z:1=A 22<@ vqt ,E* K&|4Q7I%K&|4Q7I;& !**   )!  <( L   7# LI Hs<R c `tjdk(rdnd} ||z} |dk7ritj| dddt|dz|} tj | |dz | } tj ||d z | } n|} |d k(r7tjtj| ||d z| } d}nd}tj|||dt|dz| } tj | |dz | } tj ||dz | } d| cxkrdkrnntdt|| z}tj|dz | }| dk(r| ddf}ndd| f}tj||dz |}tj|dd|t|dz|}tj| dddt|dz|}tj| |g|dz } tj|dddt|dz| } tj | |dz | } | rL|dk(rG||k(rB|dkDr!tj |d|dz | } tj"| |g|d!z } | S)"aAn inverted residual block. Args: inputs: input tensor. activation: activation function. drop_rate: float between 0 and 1, fraction of the input units to drop. name: string, block label. filters_in: integer, the number of input filters. filters_out: integer, the number of output filters. kernel_size: integer, the dimension of the convolution window. strides: integer, the stride of the convolution. expand_ratio: integer, scaling coefficient for the input filters. se_ratio: float between 0 and 1, fraction to squeeze the input filters. id_skip: boolean. Returns: output tensor for the block. rLr"r#rrF expand_convrt expand_bnrgexpand_activationrir. dwconv_padr_rbdwconv)r,r`rddepthwise_initializerrabnrzr se_squeeze se_reshape se_reduce)r`rzrerasigmoid se_expand se_excite project_conv project_bn)Nr#r#r#drop) noise_shaperaadd)rrrrrrrrrrDepthwiseConv2DrNrOrReshapemultiplyrr)rrz drop_raterar(r)r&r,r*r-r+rrPrconv_pad filters_sesese_shapes rTroros>,,./AaqG<'Gq FMM  6 %     MF % %7 9K LQ O JF  jt6I/I J1 M !| F "..q+> $    5 H_   A B!!wTD[A!DA?*4,+>?BA 8qC X 567 DV * *|0C DQ G a<AH1gH ?V^^H4,+> ? C V]]  !6 #    V]]   6 #    OOQG$*< =   2 N "   A J!!wTL5HI!LA7ars -1')&H        ;!J       {GT$*$!>B,    ##eT    p f8+   #   68+   #   68+   #   68+   #   68+   #   68+   #   68+   #   68+   #   6(..4D.E'..4D.E'..4D.E'..4D.E'..4D.E'..4D.E'..4D.E'..4D.E@A B *BC=D=,>>FFrV