o ={cA@sdZddlZddlZddlZddlmZddlmZddlm Z ddl m Z da dZ dZd Zd Zd Zd Zd ZdZe dd#ddZejedede_e dd$ddZddZddZ d%ddZdd Zd!d"ZdS)&z@Utilities for ImageNet data preprocessing & prediction decoding.N) activations)backend) data_utils) keras_exportzUhttps://storage.googleapis.com/download.tensorflow.org/data/imagenet_class_index.jsonat Preprocesses a tensor or Numpy array encoding a batch of images. Usage example with `applications.MobileNet`: ```python i = tf.keras.layers.Input([None, None, 3], dtype = tf.uint8) x = tf.cast(i, tf.float32) x = tf.keras.applications.mobilenet.preprocess_input(x) core = tf.keras.applications.MobileNet() x = core(x) model = tf.keras.Model(inputs=[i], outputs=[x]) image = tf.image.decode_png(tf.io.read_file('file.png')) result = model(image) ``` Args: x: A floating point `numpy.array` or a `tf.Tensor`, 3D or 4D with 3 color channels, with values in the range [0, 255]. The preprocessed data are written over the input data if the data types are compatible. To avoid this behaviour, `numpy.copy(x)` can be used. 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: Preprocessed `numpy.array` or a `tf.Tensor` with type `float32`. {ret} Raises: {error} a mode: One of "caffe", "tf" or "torch". Defaults to "caffe". - caffe: will convert the images from RGB to BGR, then will zero-center each color channel with respect to the ImageNet dataset, without scaling. - tf: will scale pixels between -1 and 1, sample-wise. - torch: will scale pixels between 0 and 1 and then will normalize each channel with respect to the ImageNet dataset. zE ValueError: In case of unknown `mode` or `data_format` argument.z; ValueError: In case of unknown `data_format` argument.zH The inputs pixel values are scaled between -1 and 1, sample-wise.z The input pixels values are scaled between 0 and 1 and each channel is normalized with respect to the ImageNet dataset.z The images are converted from RGB to BGR, then each color channel is zero-centered with respect to the ImageNet dataset, without scaling.z2keras.applications.imagenet_utils.preprocess_inputcaffecCsf|dvr td||durt}n |dvrtd|t|tjr,t|||dSt|||dS)z@Preprocesses a tensor or Numpy array encoding a batch of images.>torchrtfzDExpected mode to be one of `caffe`, `tf` or `torch`. Received: mode=N>Z channels_lastchannels_firstz]Expected data_format to be one of `channels_first` or `channels_last`. Received: data_format=) data_formatmode) ValueErrorrimage_data_format isinstancenpZndarray_preprocess_numpy_input_preprocess_symbolic_input)xr r rAlib/python3.10/site-packages/keras/applications/imagenet_utils.pypreprocess_inputgs"  r)r reterrorz4keras.applications.imagenet_utils.decode_predictionscst|jdks|jddkrtdt|jtdur=tjdtddd }t| }t |aWdn1s8wYg}|D]' | dddd }fd d |D}|j d ddd| |qA|S)aDecodes the prediction of an ImageNet model. Args: preds: Numpy array encoding a batch of predictions. top: Integer, how many top-guesses to return. Defaults to 5. Returns: A list of lists of top class prediction tuples `(class_name, class_description, score)`. One list of tuples per sample in batch input. Raises: ValueError: In case of invalid shape of the `pred` array (must be 2D). izx`decode_predictions` expects a batch of predictions (i.e. a 2D array of shape (samples, 1000)). Found array with shape: Nzimagenet_class_index.jsonZmodelsZ c2c37ea517e94d9795004a39431a14cb)Z cache_subdirZ file_hashcs&g|]}ttt||fqSr)tuple CLASS_INDEXstr).0iZpredrr s&z&decode_predictions..cSs|dS)Nrr)rrrrsz$decode_predictions..T)keyreverse)lenshaper rrrZget_fileCLASS_INDEX_PATHopenjsonloadZargsortsortappend)ZpredstopZfpathfresultsZ top_indicesresultrr"rdecode_predictionss.   r3cCs.t|jjtjs|jtdd}|dkr|d}|d8}|S|dkr0|d}gd}gd }n/|d krP|jd krC|d d d df}n|d d d d d df}n |dd d d f}gd}d }|d krT|jd kr|dd d d d f|d8<|dd d d d f|d8<|dd d d d f|d8<|d ur|dd d d d f|d<|dd d d d f|d<|dd d d d f|d<|S|d d dd d d d f|d8<|d d dd d d d f|d8<|d d dd d d d f|d8<|d urR|d d dd d d d f|d<|d d dd d d d f|d<|d d dd d d d f|d<|S|d|d8<|d|d8<|d|d8<|d ur|d|d<|d|d<|d|d<|S)aPreprocesses a Numpy array encoding a batch of images. Args: x: Input array, 3D or 4D. data_format: Data format of the image array. mode: One of "caffe", "tf" or "torch". - caffe: will convert the images from RGB to BGR, then will zero-center each color channel with respect to the ImageNet dataset, without scaling. - tf: will scale pixels between -1 and 1, sample-wise. - torch: will scale pixels between 0 and 1 and then will normalize each channel with respect to the ImageNet dataset. Returns: Preprocessed Numpy array. F)copyr_@?ro@g ףp= ?gv/?gCl?gZd;O?gy&1?g?r Nr.gjtY@g`"1]@gQ^@rrr).r).r).r) issubclassdtypetyperZfloatingZastyperZfloatxndim)rr r meanstdrrrrsV    """"""((( (((  rcCs6|dkr|d}|d8}|S|dkr|d}gd}gd}n1|dkrAt|d kr4|d d d d f}n|d d d d d d f}n |d d d d f}gd }d }tt| }t|t|krstj|t|t||d}nt|||}|d urtjt|t|d}|dkrt|d}||}|S)aPreprocesses a tensor encoding a batch of images. Args: x: Input tensor, 3D or 4D. data_format: Data format of the image tensor. mode: One of "caffe", "tf" or "torch". - caffe: will convert the images from RGB to BGR, then will zero-center each color channel with respect to the ImageNet dataset, without scaling. - tf: will scale pixels between -1 and 1, sample-wise. - torch: will scale pixels between 0 and 1 and then will normalize each channel with respect to the ImageNet dataset. Returns: Preprocessed tensor. rr5r6rr7r8r9r r:Nr.r;)r )r=)rrr) rr?ZconstantrZarrayr=Zbias_addcastZreshape)rr r r@rAZ mean_tensorZ std_tensorrrrrs:  rcCsT|dkrJ|rJt|dkrJ|dkr-|ddvr%tjdt|dddd |d||f}n,|d dvrBtjdt|d ddd |||d f}n|dkrTd||f}n||df}|dkrs|rs|d urq||krqtd |d ||S|r |dkr|d urt|dkrtd|ddkr|dkrtd|d|dd ur|d|ks|dd ur|d|krtd|d|d|nY|d urt|dkrtd|d dkr|dkrtd|d|dd ur|d|ks|dd ur|d|krtd|d|d|n|r|}n |dkrd}nd}|r(d |vr(td||S)aInternal utility to compute/validate a model's input shape. Args: input_shape: Either None (will return the default network input shape), or a user-provided shape to be validated. default_size: Default input width/height for the model. min_size: Minimum input width/height accepted by the model. data_format: Image data format to use. require_flatten: Whether the model is expected to be linked to a classifier via a Flatten layer. weights: One of `None` (random initialization) or 'imagenet' (pre-training on ImageNet). If weights='imagenet' input channels must be equal to 3. Returns: An integer shape tuple (may include None entries). Raises: ValueError: In case of invalid argument values. Zimagenetr:r r>rr:z]This model usually expects 1 or 3 input channels. However, it was passed an input_shape with z input channels.r) stacklevelrNzXWhen setting `include_top=True` and loading `imagenet` weights, `input_shape` should be z. Received: input_shape=z0`input_shape` must be a tuple of three integers.z6The input must have 3 channels; Received `input_shape=`rzInput size must be at least rz; Received: input_shape=)r:NN)NNr:z[If `include_top` is True, you should specify a static `input_shape`. Received: input_shape=)r'warningswarnrr )Z input_shapeZ default_sizeZmin_sizer Zrequire_flattenweightsZ default_shaperrrobtain_input_shape6s            rHcCstdkrdnd}t|||d}t|tr||f}|ddur'd}nd|ddd|ddf}|dd|ddf}|d|d|df|d|d|dffS)zReturns a tuple for zero-padding for 2D convolution with downsampling. Args: inputs: Input tensor. kernel_size: An integer or tuple/list of 2 integers. Returns: A tuple. r rrrN)rr)rr Z int_shaperint)ZinputsZ kernel_sizeZimg_dimZ input_sizeZadjustZcorrectrrr correct_pads    rJcCs@|durdSt|}|tdtdhvrtd|dS)a@validates that the classifer_activation is compatible with the weights. Args: classifier_activation: str or callable activation function weights: The pretrained weights to load. Raises: ValueError: if an activation other than `None` or `softmax` are used with pretrained weights. NZsoftmaxzOnly `None` and `softmax` activations are allowed for the `classifier_activation` argument when using pretrained weights, with `include_top=True`; Received: classifier_activation=)rgetr )Zclassifier_activationrGrrrvalidate_activations  rL)Nr)r)N)__doc__r+rEZnumpyrZkerasrrZ keras.utilsrZ tensorflow.python.util.tf_exportrrr)ZPREPROCESS_INPUT_DOCZPREPROCESS_INPUT_MODE_DOCZ"PREPROCESS_INPUT_DEFAULT_ERROR_DOCZPREPROCESS_INPUT_ERROR_DOCZPREPROCESS_INPUT_RET_DOC_TFZPREPROCESS_INPUT_RET_DOC_TORCHZPREPROCESS_INPUT_RET_DOC_CAFFErformatr3rrrHrJrLrrrrsB    $   +IB y