a Sicf@sdZddlZddlZddlmmZddlmZddl m Z ddl m Z ddl mZddlmZddlmZdd lmZd Zd d d ddddddZddddddddddddddddddddddddddddddd dddddddd d!ddddddd"d!d#dddddddd#d$dddddgZd%d&d'd(d)d*Zd%d+d'd,d)d*ZeZd-Zgd.Zddd7d8Zded;d<Zed=d>dfd?d@ZedAdBdgdCdDZedEdFdhdGdHZ edIdJdidKdLZ!edMdNdjdOdPZ"edQdRdkdSdTZ#edUdVdldWdXZ$edYdZdmd[d\Z%ej&d@d]e_ej&dDd]e_ej&dHd]e _ej&dLd]e!_ej&dPd]e"_ej&dTd]e#_ej&dXd]e$_ej&d\d]e%_ed^dnd_d`Z'edadodbdcZ(e j(je(_dS)pzEfficientNet models for Keras. Reference: - [EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks]( https://arxiv.org/abs/1905.11946) (ICML 2019) N)backend)imagenet_utils)training)VersionAwareLayers) data_utils) layer_utils) keras_exportz2https://storage.googleapis.com/keras-applications/)Z 902e53a9f72be733fc0bcb005b3ebbacZ 50bc09e76180e00e4465e1a485ddc09d)Z 1d254153d4ab51201f1646940f018540Z 74c4e6b3e1f6a1eea24c589628592432)Z b15cce36ff4dcbd00b6dd88e7857a6adZ 111f8e2ac8aa800a7a99e3239f7bfb39)Z ffd1fdc53d0ce67064dc6a9c7960ede0Z af6d107764bb5b1abb91932881670226)Z 18c95ad55216b8f92d7e70b3a046e2fcZ ebc24e6d6c33eaebbd558eafbeedf1ba)Z ace28f2a6363774853a83a0b21b9421aZ 38879255a25d3c92d5e44e04ae6cec6f)Z 165f6e37dce68623721b423839de8be5Z 9ecce42647a20130c1f39a5d4cb75743)Z 8c03f828fec3ef71311cd463b6759d99Z cbcfe4450ddf6f3ad90b1b398090fe4a)b0b1b2Zb3Zb4Zb5Zb6Zb7 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 `tf.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. Defaults to 1000 (number of ImageNet classes). 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"`. Returns: A `keras.Model` instance. )gZd;O?gy&1?g?皙?swishdefault efficientnetimagenetsoftmaxc s|dkr t}| dvs*tjj| s*td| dkrF| rF|dkrFtdtj| |dt | | d} | d urvt j | d }nt | st j | | d }n| }t d krd nd}|ffdd }fdd|}t d|}t j|d|}| dkr t dtjt|}t jt|d dd|}t j|dd dddtdd|}t j|dd|}t j|dd |}t|}d!}ttfd"d#|D}t|D]\}}|d$d!ksJ||d%|d%<||d&|d&<t|d$D]`}|d!krd|d'<|d&|d%<t|||||fd(d) |dt!|d*i|}|d7}qqt j|d+dd,dtd-d.|}t j|d/d|}t j|d0d |}| rt j"d1d |}|d!krt j#|d2d |}t$|| t j%||t&d3d4|}n6| d5krt j"d1d |}n| d6krt j'd7d |}| d ur.t()| }n|}t*j+|||d }| dkr| rld8}t,|d9d d!}nd:}t,|d9d d}||}t-j.|t/|d;|d<}|0|n| d ur|0| |S)=aT Instantiates the EfficientNet 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. 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()`) 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 `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.>Nr0zThe `weights` argument should be either `None` (random initialization), `imagenet` (pre-training on ImageNet), or the path to the weights file to be loaded.r0r1zWIf using `weights` as `"imagenet"` with `include_top` as true, `classes` should be 1000r) default_sizemin_size data_formatrequire_flattenweightsN)shape)tensorr8 channels_lastr r csB|9}t|t||d||}|d|kr:||7}t|S)z2Round number of filters based on depth multiplier.rg?)maxint)filtersdivisorZ new_filters)width_coefficient[/var/www/html/django/DPS/env/lib/python3.9/site-packages/keras/applications/efficientnet.py round_filtersTs z#EfficientNet..round_filterscstt|S)z2Round number of repeats based on depth multiplier.)r<mathceil)r)depth_coefficientr@rA round_repeats_sz#EfficientNet..round_repeatsgp?)axis?Z stem_conv_padpaddingnamervalidFZ stem_conv)rrJuse_biaskernel_initializerrKZstem_bnrGrKZstem_activationrKrc3s|]}|dVqdS)rNr@).0args)rFr@rA zEfficientNet..rrrrrKz block{}{}_aisameZtop_convrJrMrNrKZtop_bnZtop_activationavg_poolZ top_dropout predictions) activationrNrKavgr;max_poolz.h5z _notop.h5models) cache_subdir file_hash)1DEFAULT_BLOCKS_ARGStfiogfileexists ValueErrorrobtain_input_shaperimage_data_formatlayersInputis_keras_tensor Rescaling NormalizationrCsqrtIMAGENET_STDDEV_RGB ZeroPadding2D correct_padConv2DCONV_KERNEL_INITIALIZERBatchNormalization Activationcopydeepcopyfloatsum enumeraterangepopblockformatchrGlobalAveragePooling2DDropoutvalidate_activationDenseDENSE_KERNEL_INITIALIZERGlobalMaxPooling2Drget_source_inputsrModelWEIGHTS_HASHESrget_fileBASE_WEIGHTS_PATH load_weights)r?rEr3Z dropout_rateZdrop_connect_rateZ depth_divisorrZZ blocks_args model_name include_topr7 input_tensor input_shapepoolingclassesclassifier_activation img_inputbn_axisrBxbblocksirRjinputsmodel file_suffixr` file_name weights_pathr@)rErFr?rA EfficientNetsD                    rc CsBtdkrdnd} ||} |dkrptj| dddt|dd|} tj| |dd | } tj||d d | } n|} |d krtjt | ||d d| } d}nd}tj |||dt|dd| } tj| |dd | } tj||dd | } d| krdkrnnt dt || }tj |dd | }| dkrF| ddf}n dd| f}tj||dd |}tj|dd|t|dd|}tj| dddt|dd|}tj| |g|dd } tj|dddt|dd| } tj| |dd | } | r>|dkr>||kr>|dkr(tj|d|dd | } tj| |g|d!d } | 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. r:r r rVFZ expand_convrWZ expand_bnrOZexpand_activationrPrZ dwconv_padrIrLZdwconv)rrJrMdepthwise_initializerrKbnrZrZ se_squeezeZ se_reshapeZ se_reduce)rJrZrNrKsigmoidZ se_expandZ se_exciteZ project_convZ project_bn)Nr r r drop) noise_shaperKadd)rrhrirrrsrtrurprrqDepthwiseConv2Dr;r<rReshapemultiplyrr)rrZZ drop_raterKrrrrrrrrr=rZconv_padZ filters_seseZse_shaper@r@rAr}s      r}z.keras.applications.efficientnet.EfficientNetB0z!keras.applications.EfficientNetB0c Ks tdd|||||||d|S)N)rHrHr+Zefficientnetb0rrr7rrrrrrrr7rrrrrkwargsr@r@rAEfficientNetB0Hs rz.keras.applications.efficientnet.EfficientNetB1z!keras.applications.EfficientNetB1c Ks tdd|||||||d|S)N)rH皙?r+Zefficientnetb1rrrr@r@rAEfficientNetB1gs rz.keras.applications.efficientnet.EfficientNetB2z!keras.applications.EfficientNetB2c Ks tdd|||||||d|S)N)r333333?i333333?Zefficientnetb2rrrr@r@rAEfficientNetB2s rz.keras.applications.efficientnet.EfficientNetB3z!keras.applications.EfficientNetB3c Ks tdd|||||||d|S)N)rffffff?i,rZefficientnetb3rrrr@r@rAEfficientNetB3s rz.keras.applications.efficientnet.EfficientNetB4z!keras.applications.EfficientNetB4c Ks tdd|||||||d|S)N)r?i|皙?Zefficientnetb4rrrr@r@rAEfficientNetB4s rz.keras.applications.efficientnet.EfficientNetB5z!keras.applications.EfficientNetB5c Ks tdd|||||||d|S)N)g?g@irZefficientnetb5rrrr@r@rAEfficientNetB5s rz.keras.applications.efficientnet.EfficientNetB6z!keras.applications.EfficientNetB6c Ks tdd|||||||d|S)N)rg@i?Zefficientnetb6rrrr@r@rAEfficientNetB6s rz.keras.applications.efficientnet.EfficientNetB7z!keras.applications.EfficientNetB7c Ks tdd|||||||d|S)N)r"g@iXrZefficientnetb7rrrr@r@rAEfficientNetB7!s rrPz0keras.applications.efficientnet.preprocess_inputcCs|S)aA placeholder method for backward compatibility. The preprocessing logic has been included in the efficientnet 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`. r@)rr5r@r@rApreprocess_inputJsrz2keras.applications.efficientnet.decode_predictionscCstj||dS)N)top)rdecode_predictions)predsrr@r@rAr`sr) r+r+r,r-r.r/Tr0NNNr1r2) r-rrrrr r r rT)Tr0NNNr1r2)Tr0NNNr1r2)Tr0NNNr1r2)Tr0NNNr1r2)Tr0NNNr1r2)Tr0NNNr1r2)Tr0NNNr1r2)Tr0NNNr1r2)N)r))__doc__rvrCZtensorflow.compat.v2compatv2rbkerasrZkeras.applicationsr keras.enginer keras.layersr keras.utilsrr tensorflow.python.util.tf_exportrrrrarsrriBASE_DOCSTRINGrorr}rrrrrrrrr~rrr@r@r@rAs       %      J  ? m s