a Sic@sdZddlZddlZddlmmZddlmZddlm Z ddl m Z ddl m Z ddlmZddlmZdd lmZd Zd d d dddddZddddddddddddddddddddddddddddddddddddddddd d!dddddd!dd"d#dddgddddddddddd$ddddddddd$dd%ddddddd&d%d!ddddddd'd!d(ddddddd)d(d*ddddddd$d*d+dddddgddd,d,ddddddd&d,dddddddd&dd-ddddddd.d-d/ddddddd0d/d1ddddddd2d1d3ddddddd&d3d4dddddgddd,d5dddddddd5d,dddddddd,dddddddddd-ddddddd$d-d6ddddddd7d6d/dddddgddd,d5dddddddd5d,dddddddd,dddddddddd-ddddddd$d-d6ddddddd7d6d/dddddgddd,d5dddddddd5d,dddddddd,dddddddddd-ddddddd$d-d6ddddddd7d6d/dddddgddd,d5dddddddd5d,dddddddd,dddddddddd-ddddddd$d-d6ddddddd7d6d/dddddgd8Zd9d:d;dZd9d?d;d@d=d>ZdAZdBdCZdDdEZdzeeedIdJdKZd{eeedIdLdMZd|dUdVZedWdXd}dYdZZ ed[d\d~d]d^Z!ed_d`ddadbZ"edcddddedfZ#edgdhddidjZ$edkdlddmdnZ%edodpddqdrZ&ej'dZdse _ej'd^dse!_ej'dbdse"_ej'dfdse#_ej'djdse$_ej'dndse%_ej'drdse&_edtddudvZ(edwddxdyZ)e j)je)_dS)zEfficientNet V2 models for Keras. Reference: - [EfficientNetV2: Smaller Models and Faster Training]( https://arxiv.org/abs/2104.00298) (ICML 2021) N)backend)layers)imagenet_utils)training) data_utils) layer_utils) keras_exportzMhttps://storage.googleapis.com/tensorflow/keras-applications/efficientnet_v2/)Z 21ecbf6da12460d5c40bb2f29ceb2188Z 893217f2bb855e2983157299931e43ff)Z 069f0534ff22adf035c89e2d9547a9dcZ 0e80663031ca32d657f9caa404b6ec37)Z 424e49f28180edbde1e94797771950a7Z 1dfe2e7a5d45b6632553a8961ea609eb)Z 1f1fc43bd98a6e4fd8fdfd551e02c7a0Z f6abf7b5849ac99a89b50dd3fd532856)Z e1d88a8495beba45748fedd0cecbe016Z af0682fb74e8c54910f2d4393339c070)Z a3bf6aa3276309f4fc6a34aa114c95cdZ 1b8dc055df72dde80d614482840fe342)Z 27e6d408b53c7ebc868fefa357689935Z b0b66b5c863aef5b46e8608fe1711615)b0b1b2b3z-sz-mz-l) kernel_size num_repeat input_filtersoutput_filters expand_ratiose_ratiostrides conv_type0@)rrrrrrrrg? Pi0i ` ip)efficientnetv2-sefficientnetv2-mefficientnetv2-lefficientnetv2-b0efficientnetv2-b1efficientnetv2-b2efficientnetv2-b3VarianceScalingg@fan_outtruncated_normal)scalemode distribution) class_nameconfiggUUUUUU?uniforma Instantiates the {name} architecture. Reference: - [EfficientNetV2: Smaller Models and Faster Training]( https://arxiv.org/abs/2104.00298) (ICML 2021) 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 EfficientNetV2, by default input preprocessing is included as a part of the model (as a `Rescaling` layer), and thus `tf.keras.applications.efficientnet_v2.preprocess_input` is actually a pass-through function. In this use case, EfficientNetV2 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 EfficientNetV2 models expect their inputs to be float tensors of pixels with values in the [-1, 1] range. Args: include_top: Boolean, 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. Defaults to 1000 (number of ImageNet classes). classifier_activation: A string 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"`. Returns: A `keras.Model` instance. cCs6||9}|p|}t|t||d||}t|S)z2Round number of filters based on depth multiplier.r)maxint)filterswidth_coefficient min_depth depth_divisorZ minimum_depth new_filtersrK^/var/www/html/django/DPS/env/lib/python3.9/site-packages/keras/applications/efficientnet_v2.py round_filtersfsrMcCstt||S)z2Round number of repeats based on depth multiplier.)rEmathceilrepeatsdepth_coefficientrKrKrL round_repeatsqsrS?swish皙?)rrsurvival_probabilityc sJtdkrdnddur&td f dd} | S)z2MBConv block: Mobile Inverted Residual Bottleneck. channels_lastr rNblock0c s}dkrbtj|ddtddddd|}tjdd|}tjd d |}n|}tj tdddd d |}tjd d|}tjdd |}dkrdkrnn0tdt}tjdd |}dkr |ddf}n dd|f}tj |dd |}tj|ddtdd|}tj|dddtdd|}tj ||gdd }tjddtddddd|}tjdd|} dkrkr rtj ddd|}tj ||gdd }|S)NrsamerXF expand_convrFrrkernel_initializerpadding data_formatuse_biasname expand_bnaxismomentumraexpand_activationraZdwconv2)rrdepthwise_initializerr^r_r`rabn activationr se_squeeze se_reshape se_reducer^rjr]rasigmoid se_expand se_excite project_conv project_bnNrrrdrop noise_shaperaadd) rConv2DCONV_KERNEL_INITIALIZERBatchNormalization ActivationDepthwiseConv2DrDrEGlobalAveragePooling2DReshapemultiplyDropoutrxinputsrFx filters_sesese_shape rjbn_axis bn_momentumrrrrarrrrWrKrLapplys           zMBConvBlock..applyrimage_data_formatget_uid rrrrrrrrjrWrarrKrrL MBConvBlockvs   ^rc sJtdkrdnddur&td f dd} | S)zQFused MBConv Block: Fusing the proj conv1x1 and depthwise_conv into a conv2d.rXr rNrYc s}dkrbtj| tddddd|}tjdd|}tjd d |}n|}d kr|dkr,nntdt}tjd d |}dkr|ddf}n dd|f}tj|dd |}tj|ddtdd|}tj|dddtdd|}tj ||gdd }tjdkr@dndkrPdn tdddd|}tjdd|}dkrtjdd |} dkrkr rtj ddd|}tj ||gdd }|S)NrrXrZFr[)rrr]r_r^r`rarbrcrfrjrarrkrgrlrmrnrorprqrr)rrr]r^r`rarsZproject_activationrtrurvrx) rryrzr{r|rDrEr~rrrrxrrrKrLrs         zFusedMBConvBlock..applyrrrKrrLFusedMBConvBlocks   Qr皙?defaultefficientnetv2Timagenetsoftmaxc& Cs,| dkrt| } | dvs4tjj| s4td| | dkrV| rV|dkrVtd|tj||dt | | d}| d urt j |d }nt | st j | |d }n| }t d krd nd}|}|r.||d}| dddr|d krt jdd|}t jgdgd|d|}nt jddd|}t| dd|||d}t j|d dtdddd |}t j||d!d"|}t j|d#d$|}t| } d}ttd%d&| D}t| D]\}}|d'dksJt|d|||d|d<t|d(|||d|d(<ttd)|d*}t|d'|d+}t|D]d}|dkrNd|d,<|d(|d<|f|||||d- |dt!|d.d/||}|d7}q,qtd0|||d}t j|ddtdd dd1d2|}t j||d3d"|}t j|d4d5|}| rHt j"d6d$|}|dkrt j#|d7d$|}t$|| t j%||t&t'dd8d9|}n6|d:krdt j"d6d$|}n|d;kr~t j(dd d}#nd?}"t-| d>d d}#| |"}$t.j/|$t0|$d@|#dA}%|!1|%n| d ur(|!1| |!S)Ba Instantiates the EfficientNetV2 architecture using given scaling coefficients. 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. min_depth: integer, minimum number of filters. bn_momentum: float. Momentum parameter for Batch Normalization layers. activation: activation function. blocks_args: list of dicts, parameters to construct block modules. model_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()`) or numpy array 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 string 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. Defaults to `True`. Returns: A `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. r>NrzThe `weights` argument should be either `None` (random initialization), `imagenet` (pre-training on ImageNet), or the path to the weights file to be loaded.Received: weights=rrziIf using `weights` as `'imagenet'` with `include_top` as true, `classes` should be 1000Received: classes=r)) default_sizemin_sizer_require_flattenweightsN)shape)tensorrrXr r-bgp?)r>)g ףp= ?gv/?gCl?)gg٪?g.?gQ?)meanvariancerdg?)r>offsetrr)rFrGrHrIrrZF stem_conv)rFrrr]r^r`rastem_bnrcstem_activationrgcss|]}|dVqdS)rNrK).0argsrKrKrL z!EfficientNetV2..rr)rrrrPrz block{}{}_a)rjrrWraitop_convr\top_bntop_activationravg_pool top_dropout predictions)rjr]bias_initializerraavgrDmax_poolz.h5z _notop.h5models) cache_subdir file_hash)2DEFAULT_BLOCKS_ARGStfiogfileexists ValueErrorrobtain_input_shaperrrInputis_keras_tensorsplit startswith Rescaling NormalizationrMryrzr{r|copydeepcopyfloatsum enumeraterrpoprSrangeformatchrr~rvalidate_activationDenseDENSE_KERNEL_INITIALIZERconstant_initializerGlobalMaxPooling2Drget_source_inputsrModelWEIGHTS_HASHESrget_fileBASE_WEIGHTS_PATH load_weights)&rGrRr dropout_ratedrop_connect_raterIrHrrj blocks_args model_name include_topr input_tensor input_shapepoolingclassesclassifier_activationinclude_preprocessing img_inputrr num_channelsZ stem_filtersrblocksirblockrQjZ top_filtersrmodel file_suffixr file_name weights_pathrKrKrLEfficientNetV2PsNG                          rz3keras.applications.efficientnet_v2.EfficientNetV2B0z#keras.applications.EfficientNetV2B0cCs tdddd||||||||d S)N?r.r7 rGrRrrrrrrrrrrrrrrrrrrrrKrKrLEfficientNetV2B0Zsrz3keras.applications.efficientnet_v2.EfficientNetV2B1z#keras.applications.EfficientNetV2B1cCs tdddd||||||||d S)Nr皙?r8rrrrKrKrLEfficientNetV2B1xsrz3keras.applications.efficientnet_v2.EfficientNetV2B2z#keras.applications.EfficientNetV2B2cCs tdddd||||||||d S)Nr333333?ir9rrrrKrKrLEfficientNetV2B2srz3keras.applications.efficientnet_v2.EfficientNetV2B3z#keras.applications.EfficientNetV2B3cCs tdddd||||||||d S)Nrgffffff?i,r:rrrrKrKrLEfficientNetV2B3srz2keras.applications.efficientnet_v2.EfficientNetV2Sz"keras.applications.EfficientNetV2ScCs tdddd||||||||d S)Nrr0r4rrrrKrKrLEfficientNetV2Ssrz2keras.applications.efficientnet_v2.EfficientNetV2Mz"keras.applications.EfficientNetV2McCs tdddd||||||||d S)Nrr5rrrrKrKrLEfficientNetV2Msrz2keras.applications.efficientnet_v2.EfficientNetV2Lz"keras.applications.EfficientNetV2LcCs tdddd||||||||d S)Nrrr6rrrrKrKrLEfficientNetV2Lsrrgz3keras.applications.efficientnet_v2.preprocess_inputcCs|S)aA placeholder method for backward compatibility. The preprocessing logic has been included in the EfficientNetV2 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. Defaults to None, in which case the global setting `tf.keras.backend.image_data_format()` is used (unless you changed it, it defaults to "channels_last").{mode} Returns: Unchanged `numpy.array` or `tf.Tensor`. rK)rr_rKrKrLpreprocess_input5srz5keras.applications.efficientnet_v2.decode_predictionscCstj||dS)N)top)rdecode_predictions)predsrrKrKrLrKsr)rr rrrTrUrVN)rr rrrTrUrVN)rrr3r3rTrUrrTrNNNrrT)TrNNNrrT)TrNNNrrT)TrNNNrrT)TrNNNrrT)TrNNNrrT)TrNNNrrT)TrNNNrrT)N)r#)*__doc__rrNZtensorflow.compat.v2compatv2rkerasrrZkeras.applicationsr keras.enginer keras.utilsrr tensorflow.python.util.tf_exportrrrrrzrBASE_DOCSTRINGrMrSrErrrrrrrrrrrrrrrKrKrKrLs       "     @      J      J     @     @     @     M  C  v k