:py:mod:`neural_compressor.data.transforms.imagenet_transform`
==============================================================

.. py:module:: neural_compressor.data.transforms.imagenet_transform

.. autoapi-nested-parse::

   Neural Compressor built-in imagenet transforms.



Module Contents
---------------

Classes
~~~~~~~

.. autoapisummary::

   neural_compressor.data.transforms.imagenet_transform.QuantizedInput
   neural_compressor.data.transforms.imagenet_transform.LabelShift
   neural_compressor.data.transforms.imagenet_transform.ParseDecodeImagenet
   neural_compressor.data.transforms.imagenet_transform.ParseDecodeImagenetTransform
   neural_compressor.data.transforms.imagenet_transform.TensorflowTransposeLastChannel
   neural_compressor.data.transforms.imagenet_transform.TensorflowShiftRescale
   neural_compressor.data.transforms.imagenet_transform.TensorflowResizeCropImagenetTransform
   neural_compressor.data.transforms.imagenet_transform.BilinearImagenetTransform
   neural_compressor.data.transforms.imagenet_transform.OnnxBilinearImagenetTransform
   neural_compressor.data.transforms.imagenet_transform.ONNXResizeCropImagenetTransform
   neural_compressor.data.transforms.imagenet_transform.ResizeWithAspectRatio




.. py:class:: QuantizedInput(dtype, scale=None)




   Convert the dtype of input to quantize it.

   :param dtype: desired image dtype, support 'uint8', 'int8'
   :type dtype: str
   :param scale: scaling ratio of each point in image
   :type scale: float, default=None

   :returns: tuple of processed image and label


.. py:class:: LabelShift(label_shift=0)




   Convert label to label - label_shift.

   :param label_shift: number of label shift
   :type label_shift: int, default=0

   :returns: tuple of processed image and label


.. py:class:: ParseDecodeImagenet


   Parse features in Example proto.

   :returns: tuple of parsed image and label


.. py:class:: ParseDecodeImagenetTransform




   Imagenet decoding will be performed automatically from Neural Compressor v1.4.

   :returns: sample


.. py:class:: TensorflowTransposeLastChannel




   Transpose NHWC to NCHW.

   :returns: tuple of processed image and label


.. py:class:: TensorflowShiftRescale




   Label shift by 1 and rescale.

   :returns: tuple of processed image and label


.. py:class:: TensorflowResizeCropImagenetTransform(height, width, random_crop=False, resize_side=256, resize_method='bilinear', random_flip_left_right=False, mean_value=[0.0, 0.0, 0.0], scale=1.0, data_format='channels_last', subpixels='RGB')




   Combination of a series of transforms which is applicable to images in Imagenet.

   :param height: Height of the result
   :type height: int
   :param width: Width of the result
   :type width: int
   :param random_crop: whether to random crop
   :type random_crop: bool, default=False
   :param resize_side: desired shape after resize operation
   :type resize_side: int, default=256
   :param random_flip_left_right: whether to random flip left and right
   :type random_flip_left_right: bool, default=False
   :param mean_value: means for each channel
   :type mean_value: list, default=[0.0,0.0,0.0]
   :param scale: std value
   :type scale: float, default=1.0

   :returns: tuple of processed image and label


.. py:class:: BilinearImagenetTransform(height, width, central_fraction=0.875, mean_value=[0.0, 0.0, 0.0], scale=1.0)




   Combination of a series of transforms which is applicable to images in Imagenet.

   :param height: Height of the result
   :param width: Width of the result
   :param central_fraction: fraction of size to crop
   :type central_fraction: float, default=0.875
   :param mean_value: means for each channel
   :type mean_value: list, default=[0.0,0.0,0.0]
   :param scale: std value
   :type scale: float, default=1.0

   :returns: tuple of processed image and label


.. py:class:: OnnxBilinearImagenetTransform(height, width, central_fraction=0.875, mean_value=[0.0, 0.0, 0.0], scale=1.0)




   Combination of a series of transforms which is applicable to images in Imagenet.

   :param height: Height of the result
   :param width: Width of the result
   :param central_fraction: fraction of size to crop
   :type central_fraction: float, default=0.875
   :param mean_value: means for each channel
   :type mean_value: list, default=[0.0,0.0,0.0]
   :param scale: std value
   :type scale: float, default=1.0

   :returns: tuple of processed image and label


.. py:class:: ONNXResizeCropImagenetTransform(height, width, random_crop=False, resize_side=256, mean_value=[0.0, 0.0, 0.0], std_value=[0.229, 0.224, 0.225], resize_method='bilinear', data_format='channels_last', subpixels='RGB')




   Combination of a series of transforms which is applicable to images in Imagenet.

   :param height: Height of the result
   :param width: Width of the result
   :param central_fraction: fraction of size to crop
   :type central_fraction: float, default=0.875
   :param mean_value: means for each channel
   :type mean_value: list, default=[0.0,0.0,0.0]
   :param scale: std value
   :type scale: float, default=1.0

   :returns: tuple of processed image and label


.. py:class:: ResizeWithAspectRatio(height, width, scale=87.5, inter_pol=cv2.INTER_AREA)




   Resize the image with aspect ratio.

   :returns: image and label