minerva.losses.dice
===================

.. py:module:: minerva.losses.dice


Attributes
----------

.. autoapisummary::

   minerva.losses.dice.BINARY_MODE
   minerva.losses.dice.MULTICLASS_MODE
   minerva.losses.dice.MULTILABEL_MODE


Classes
-------

.. autoapisummary::

   minerva.losses.dice.DiceLoss
   minerva.losses.dice.MultiClassDiceCELoss


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

.. py:data:: BINARY_MODE
   :value: 'binary'


.. py:class:: DiceLoss(mode, classes = None, log_loss = False, from_logits = True, smooth = 0.0, ignore_index = None, eps = 1e-07)

   Bases: :py:obj:`torch.nn.modules.loss._Loss`


   Base class for all neural network modules.

   Your models should also subclass this class.

   Modules can also contain other Modules, allowing them to be nested in
   a tree structure. You can assign the submodules as regular attributes::

       import torch.nn as nn
       import torch.nn.functional as F


       class Model(nn.Module):
           def __init__(self) -> None:
               super().__init__()
               self.conv1 = nn.Conv2d(1, 20, 5)
               self.conv2 = nn.Conv2d(20, 20, 5)

           def forward(self, x):
               x = F.relu(self.conv1(x))
               return F.relu(self.conv2(x))

   Submodules assigned in this way will be registered, and will also have their
   parameters converted when you call :meth:`to`, etc.

   .. note::
       As per the example above, an ``__init__()`` call to the parent class
       must be made before assignment on the child.

   :ivar training: Boolean represents whether this module is in training or
                   evaluation mode.
   :vartype training: bool

   Initialize the DiceLoss class.

   Parameters
   ----------
   mode : str
       Loss mode. Valid options are 'binary', 'multiclass', or 'multilabel'.
   classes : Optional[List[int]], optional
       List of classes that contribute in loss computation. By default, all channels are included. By default None
   log_loss : bool, optional
       If True, loss is computed as `- log(dice_coeff)`. If False, loss is computed as `1 - dice_coeff`, by default False
   from_logits : bool, optional
       If True, assumes input is raw logits. If False, assumes input is probabilities., by default True
   smooth : float, optional
       Smoothness constant for dice coefficient (a), by default 0.0
   ignore_index : Optional[int], optional
       Label that indicates ignored pixels (does not contribute to loss), by default None
   eps : float, optional
       A small epsilon for numerical stability to avoid zero division error (denominator will be always greater or equal to eps), by default 1e-7

   Raises
   ------
   AssertionError
       If the mode is not one of 'binary', 'multiclass', or 'multilabel' and classes are being masked with mode='binary'.


   .. py:method:: aggregate_loss(loss)


   .. py:attribute:: classes
      :value: None



   .. py:method:: compute_score(output, target, smooth=0.0, eps=1e-07, dims=None)


   .. py:attribute:: eps
      :value: 1e-07



   .. py:method:: forward(y_pred, y_true)


   .. py:attribute:: from_logits
      :value: True



   .. py:attribute:: ignore_index
      :value: None



   .. py:attribute:: log_loss
      :value: False



   .. py:attribute:: mode


   .. py:attribute:: smooth
      :value: 0.0



.. py:data:: MULTICLASS_MODE
   :value: 'multiclass'


.. py:data:: MULTILABEL_MODE
   :value: 'multilabel'


.. py:class:: MultiClassDiceCELoss(weight_ce = 1.0, weight_dice = 1.0)

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


   Base class for all neural network modules.

   Your models should also subclass this class.

   Modules can also contain other Modules, allowing them to be nested in
   a tree structure. You can assign the submodules as regular attributes::

       import torch.nn as nn
       import torch.nn.functional as F


       class Model(nn.Module):
           def __init__(self) -> None:
               super().__init__()
               self.conv1 = nn.Conv2d(1, 20, 5)
               self.conv2 = nn.Conv2d(20, 20, 5)

           def forward(self, x):
               x = F.relu(self.conv1(x))
               return F.relu(self.conv2(x))

   Submodules assigned in this way will be registered, and will also have their
   parameters converted when you call :meth:`to`, etc.

   .. note::
       As per the example above, an ``__init__()`` call to the parent class
       must be made before assignment on the child.

   :ivar training: Boolean represents whether this module is in training or
                   evaluation mode.
   :vartype training: bool

   Combined Dice and Cross-Entropy loss for multi-class segmentation.

   Combines Dice loss for handling class imbalance and Cross-Entropy loss for
   pixel-wise classification stability, improving segmentation performance.

   Parameters
   ----------
   weight_ce : float, optional
       Weight for Cross-Entropy loss component (default is 1.0).
   weight_dice : float, optional
       Weight for Dice loss component (default is 1.0).


   .. py:method:: _dice_loss(probs, targets)

      Calculate Dice loss for all classes.

      Computes Dice loss as 1 - (2 * intersection) / (pred + target) averaged
      over all classes.

      Parameters
      ----------
      probs : torch.Tensor
          Softmax probabilities of shape (batch_size, num_classes, H, W).
      targets : torch.Tensor
          Ground truth labels of shape (batch_size, H, W).



   .. py:method:: forward(inputs, targets)

      Calculate combined Dice and Cross-Entropy loss.

      Parameters
      ----------
      inputs : torch.Tensor or list of torch.Tensor
          Model predictions. If a list, assumes deep supervision with multiple outputs.
      targets : torch.Tensor
          Ground truth segmentation masks of shape (batch_size, H, W) or
          (batch_size, 1, H, W).



   .. py:attribute:: weight_ce
      :value: 1.0



   .. py:attribute:: weight_dice
      :value: 1.0