pyiqa.losses
============

.. py:module:: pyiqa.losses


Submodules
----------

.. toctree::
   :maxdepth: 1

   /autoapi/pyiqa/losses/iqa_losses/index
   /autoapi/pyiqa/losses/loss_util/index
   /autoapi/pyiqa/losses/losses/index




Package Contents
----------------

.. py:class:: CharbonnierLoss(loss_weight=1.0, reduction='mean', eps=1e-12)

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


   Charbonnier loss (one variant of Robust L1Loss, a differentiable
   variant of L1Loss).

   Described in "Deep Laplacian Pyramid Networks for Fast and Accurate
       Super-Resolution".

   :param loss_weight: Loss weight for L1 loss. Default: 1.0.
   :type loss_weight: float
   :param reduction: Specifies the reduction to apply to the output.
                     Supported choices are 'none' | 'mean' | 'sum'. Default: 'mean'.
   :type reduction: str
   :param eps: A value used to control the curvature near zero. Default: 1e-12.
   :type eps: float


   .. py:method:: forward(pred, target, weight=None, **kwargs)

      :param pred: of shape (N, C, H, W). Predicted tensor.
      :type pred: Tensor
      :param target: of shape (N, C, H, W). Ground truth tensor.
      :type target: Tensor
      :param weight: of shape (N, C, H, W). Element-wise weights. Default: None.
      :type weight: Tensor, optional



.. py:class:: L1Loss(loss_weight=1.0, reduction='mean')

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


   L1 (mean absolute error, MAE) loss.

   :param loss_weight: Loss weight for L1 loss. Default: 1.0.
   :type loss_weight: float
   :param reduction: Specifies the reduction to apply to the output.
                     Supported choices are 'none' | 'mean' | 'sum'. Default: 'mean'.
   :type reduction: str


   .. py:method:: forward(pred, target, weight=None, **kwargs)

      :param pred: of shape (N, C, H, W). Predicted tensor.
      :type pred: Tensor
      :param target: of shape (N, C, H, W). Ground truth tensor.
      :type target: Tensor
      :param weight: of shape (N, C, H, W). Element-wise weights. Default: None.
      :type weight: Tensor, optional



.. py:class:: MSELoss(loss_weight=1.0, reduction='mean')

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


   MSE (L2) loss.

   :param loss_weight: Loss weight for MSE loss. Default: 1.0.
   :type loss_weight: float
   :param reduction: Specifies the reduction to apply to the output.
                     Supported choices are 'none' | 'mean' | 'sum'. Default: 'mean'.
   :type reduction: str


   .. py:method:: forward(pred, target, weight=None, **kwargs)

      :param pred: of shape (N, C, H, W). Predicted tensor.
      :type pred: Tensor
      :param target: of shape (N, C, H, W). Ground truth tensor.
      :type target: Tensor
      :param weight: of shape (N, C, H, W). Element-wise weights. Default: None.
      :type weight: Tensor, optional



.. py:class:: WeightedTVLoss(loss_weight=1.0, reduction='mean')

   Bases: :py:obj:`L1Loss`


   Weighted TV loss.

   :param loss_weight: Loss weight. Default: 1.0.
   :type loss_weight: float


   .. py:method:: forward(pred, weight=None)

      :param pred: of shape (N, C, H, W). Predicted tensor.
      :type pred: Tensor
      :param target: of shape (N, C, H, W). Ground truth tensor.
      :type target: Tensor
      :param weight: of shape (N, C, H, W). Element-wise weights. Default: None.
      :type weight: Tensor, optional



.. py:class:: EMDLoss(loss_weight=1.0, r=2, reduction='mean')

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


   EMD (earth mover distance) loss.


   .. py:method:: forward(pred, target, weight=None, **kwargs)


.. py:class:: PLCCLoss(loss_weight=1.0)

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


   PLCC loss, induced from Pearson’s Linear Correlation Coefficient.


   .. py:method:: forward(pred, target)


.. py:class:: NiNLoss(loss_weight=1.0, p=1, q=2)

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


   NiN (Norm in Norm) loss

   Reference:

       - Dingquan Li, Tingting Jiang, and Ming Jiang. Norm-in-Norm Loss with Faster Convergence and Better
         Performance for Image Quality Assessment. ACMM2020.
       - https://arxiv.org/abs/2008.03889
       - https://github.com/lidq92/LinearityIQA

   This loss can be simply described as: l1_norm(normalize(pred - pred_mean), normalize(target - target_mean))



   .. py:method:: forward(pred, target)