pyiqa.data.transforms
=====================

.. py:module:: pyiqa.data.transforms






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

.. py:function:: transform_mapping(key, args)

.. py:class:: PairedToTensor

   Bases: :py:obj:`torchvision.transforms.ToTensor`


   Pair version of to tensor


   .. py:method:: to_tensor(x)


.. py:class:: PairedCenterCrop

   Bases: :py:obj:`torchvision.transforms.CenterCrop`


   Pair version of center crop


   .. py:method:: forward(imgs)


.. py:class:: PairedRandomCrop

   Bases: :py:obj:`torchvision.transforms.RandomCrop`


   Pair version of random crop


   .. py:method:: forward(imgs)


.. py:class:: PairedRandomErasing

   Bases: :py:obj:`torchvision.transforms.RandomErasing`


   Pair version of random erasing


   .. py:method:: forward(imgs)


.. py:class:: PairedRandomHorizontalFlip

   Bases: :py:obj:`torchvision.transforms.RandomHorizontalFlip`


   Pair version of random hflip


   .. py:method:: forward(imgs)


.. py:class:: PairedRandomVerticalFlip

   Bases: :py:obj:`torchvision.transforms.RandomVerticalFlip`


   Pair version of random vflip


   .. py:method:: forward(imgs)


.. py:class:: PairedRandomRot90(p=0.5)

   Bases: :py:obj:`torch.nn.Module`


   Pair version of random 90 rotation


   .. py:method:: forward(imgs)


.. py:class:: PairedResize

   Bases: :py:obj:`torchvision.transforms.Resize`


   Pair version of resize


   .. py:method:: forward(imgs)


.. py:class:: PairedAdaptiveResize

   Bases: :py:obj:`torchvision.transforms.Resize`


   ARP preserved resize when necessary


   .. py:method:: forward(imgs)


.. py:class:: PairedRandomARPResize(size_range, interpolation=tf.InterpolationMode.BILINEAR, antialias=None)

   Bases: :py:obj:`torch.nn.Module`


   Pair version of resize


   .. py:method:: forward(imgs)


.. py:class:: PairedRandomSquareResize(size_range, interpolation=tf.InterpolationMode.BILINEAR, antialias=None)

   Bases: :py:obj:`torch.nn.Module`


   Pair version of resize


   .. py:method:: forward(imgs)


.. py:class:: PairedAdaptivePadding(target_size, fill=0, padding_mode='constant')

   Bases: :py:obj:`torch.nn.Module`


   Pair version of resize


   .. py:method:: get_padding(x)


   .. py:method:: forward(imgs)


.. py:function:: mod_crop(img, scale)

   Mod crop images, used during testing.

   :param img: Input image.
   :type img: ndarray
   :param scale: Scale factor.
   :type scale: int

   :returns: Result image.
   :rtype: ndarray


.. py:function:: augment(imgs, hflip=True, rotation=True, flows=None, return_status=False)

   Augment: horizontal flips OR rotate (0, 90, 180, 270 degrees).

   We use vertical flip and transpose for rotation implementation.
   All the images in the list use the same augmentation.

   :param imgs: Images to be augmented. If the input
                is an ndarray, it will be transformed to a list.
   :type imgs: list[ndarray] | ndarray
   :param hflip: Horizontal flip. Default: True.
   :type hflip: bool
   :param rotation: Ratotation. Default: True.
   :type rotation: bool
   :param flows (list[ndarray]: Flows to be augmented. If the input is an
                                ndarray, it will be transformed to a list.
                                Dimension is (h, w, 2). Default: None.
   :param return_status: Return the status of flip and rotation.
                         Default: False.
   :type return_status: bool

   :returns:

             Augmented images and flows. If returned
                 results only have one element, just return ndarray.
   :rtype: list[ndarray] | ndarray


.. py:function:: img_rotate(img, angle, center=None, scale=1.0)

   Rotate image.

   :param img: Image to be rotated.
   :type img: ndarray
   :param angle: Rotation angle in degrees. Positive values mean
                 counter-clockwise rotation.
   :type angle: float
   :param center: Rotation center. If the center is None,
                  initialize it as the center of the image. Default: None.
   :type center: tuple[int]
   :param scale: Isotropic scale factor. Default: 1.0.
   :type scale: float