[Enhance] Support reading class_weight from file in loss functions to help MMDet3D (open-mmlab#513)

* support reading class_weight from file in loss function

* add unit test of loss with class_weight from file

* minor fix

* move get_class_weight to utils

Author: Wuziyi616

mmseg/models/losses/cross_entropy_loss.py

@@ -3,7 +3,7 @@
 import torch.nn.functional as F
 
 from ..builder import LOSSES
-from .utils import weight_reduce_loss
+from .utils import get_class_weight, weight_reduce_loss
 
 
 def cross_entropy(pred,
@@ -146,8 +146,8 @@ class CrossEntropyLoss(nn.Module):
             Defaults to False.
         reduction (str, optional): . Defaults to 'mean'.
             Options are "none", "mean" and "sum".
-        class_weight (list[float], optional): Weight of each class.
-            Defaults to None.
+        class_weight (list[float] | str, optional): Weight of each class. If in
+            str format, read them from a file. Defaults to None.
         loss_weight (float, optional): Weight of the loss. Defaults to 1.0.
     """
 
@@ -163,7 +163,7 @@ def __init__(self,
         self.use_mask = use_mask
         self.reduction = reduction
         self.loss_weight = loss_weight
-        self.class_weight = class_weight
+        self.class_weight = get_class_weight(class_weight)
 
         if self.use_sigmoid:
             self.cls_criterion = binary_cross_entropy

mmseg/models/losses/dice_loss.py

@@ -5,7 +5,7 @@
 import torch.nn.functional as F
 
 from ..builder import LOSSES
-from .utils import weighted_loss
+from .utils import get_class_weight, weighted_loss
 
 
 @weighted_loss
@@ -63,8 +63,8 @@ class DiceLoss(nn.Module):
         reduction (str, optional): The method used to reduce the loss. Options
             are "none", "mean" and "sum". This parameter only works when
             per_image is True. Default: 'mean'.
-        class_weight (list[float], optional): The weight for each class.
-            Default: None.
+        class_weight (list[float] | str, optional): Weight of each class. If in
+            str format, read them from a file. Defaults to None.
         loss_weight (float, optional): Weight of the loss. Default to 1.0.
         ignore_index (int | None): The label index to be ignored. Default: 255.
     """
@@ -81,7 +81,7 @@ def __init__(self,
         self.smooth = smooth
         self.exponent = exponent
         self.reduction = reduction
-        self.class_weight = class_weight
+        self.class_weight = get_class_weight(class_weight)
         self.loss_weight = loss_weight
         self.ignore_index = ignore_index
 

mmseg/models/losses/lovasz_loss.py

@@ -8,7 +8,7 @@
 import torch.nn.functional as F
 
 from ..builder import LOSSES
-from .utils import weight_reduce_loss
+from .utils import get_class_weight, weight_reduce_loss
 
 
 def lovasz_grad(gt_sorted):
@@ -240,8 +240,8 @@ class LovaszLoss(nn.Module):
         reduction (str, optional): The method used to reduce the loss. Options
             are "none", "mean" and "sum". This parameter only works when
             per_image is True. Default: 'mean'.
-        class_weight (list[float], optional): The weight for each class.
-            Default: None.
+        class_weight (list[float] | str, optional): Weight of each class. If in
+            str format, read them from a file. Defaults to None.
         loss_weight (float, optional): Weight of the loss. Defaults to 1.0.
     """
 
@@ -269,7 +269,7 @@ def __init__(self,
         self.per_image = per_image
         self.reduction = reduction
         self.loss_weight = loss_weight
-        self.class_weight = class_weight
+        self.class_weight = get_class_weight(class_weight)
 
     def forward(self,
                 cls_score,

mmseg/models/losses/utils.py

@@ -1,8 +1,28 @@
 import functools
 
+import mmcv
+import numpy as np
 import torch.nn.functional as F
 
 
+def get_class_weight(class_weight):
+    """Get class weight for loss function.
+
+    Args:
+        class_weight (list[float] | str | None): If class_weight is a str,
+            take it as a file name and read from it.
+    """
+    if isinstance(class_weight, str):
+        # take it as a file path
+        if class_weight.endswith('.npy'):
+            class_weight = np.load(class_weight)
+        else:
+            # pkl, json or yaml
+            class_weight = mmcv.load(class_weight)
+
+    return class_weight
+
+
 def reduce_loss(loss, reduction):
     """Reduce loss as specified.
 

tests/test_models/test_losses/test_ce_loss.py

@@ -25,6 +25,34 @@ def test_ce_loss():
     fake_label = torch.Tensor([1]).long()
     assert torch.allclose(loss_cls(fake_pred, fake_label), torch.tensor(40.))
 
+    # test loss with class weights from file
+    import os
+    import tempfile
+    import mmcv
+    import numpy as np
+    tmp_file = tempfile.NamedTemporaryFile()
+
+    mmcv.dump([0.8, 0.2], f'{tmp_file.name}.pkl', 'pkl')  # from pkl file
+    loss_cls_cfg = dict(
+        type='CrossEntropyLoss',
+        use_sigmoid=False,
+        class_weight=f'{tmp_file.name}.pkl',
+        loss_weight=1.0)
+    loss_cls = build_loss(loss_cls_cfg)
+    assert torch.allclose(loss_cls(fake_pred, fake_label), torch.tensor(40.))
+
+    np.save(f'{tmp_file.name}.npy', np.array([0.8, 0.2]))  # from npy file
+    loss_cls_cfg = dict(
+        type='CrossEntropyLoss',
+        use_sigmoid=False,
+        class_weight=f'{tmp_file.name}.npy',
+        loss_weight=1.0)
+    loss_cls = build_loss(loss_cls_cfg)
+    assert torch.allclose(loss_cls(fake_pred, fake_label), torch.tensor(40.))
+    tmp_file.close()
+    os.remove(f'{tmp_file.name}.pkl')
+    os.remove(f'{tmp_file.name}.npy')
+
     loss_cls_cfg = dict(
         type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0)
     loss_cls = build_loss(loss_cls_cfg)

tests/test_models/test_losses/test_dice_loss.py

@@ -16,6 +16,36 @@ def test_dice_lose():
     labels = (torch.rand(8, 4, 4) * 3).long()
     dice_loss(logits, labels)
 
+    # test loss with class weights from file
+    import os
+    import tempfile
+    import mmcv
+    import numpy as np
+    tmp_file = tempfile.NamedTemporaryFile()
+
+    mmcv.dump([1.0, 2.0, 3.0], f'{tmp_file.name}.pkl', 'pkl')  # from pkl file
+    loss_cfg = dict(
+        type='DiceLoss',
+        reduction='none',
+        class_weight=f'{tmp_file.name}.pkl',
+        loss_weight=1.0,
+        ignore_index=1)
+    dice_loss = build_loss(loss_cfg)
+    dice_loss(logits, labels, ignore_index=None)
+
+    np.save(f'{tmp_file.name}.npy', np.array([1.0, 2.0, 3.0]))  # from npy file
+    loss_cfg = dict(
+        type='DiceLoss',
+        reduction='none',
+        class_weight=f'{tmp_file.name}.pkl',
+        loss_weight=1.0,
+        ignore_index=1)
+    dice_loss = build_loss(loss_cfg)
+    dice_loss(logits, labels, ignore_index=None)
+    tmp_file.close()
+    os.remove(f'{tmp_file.name}.pkl')
+    os.remove(f'{tmp_file.name}.npy')
+
     # test dice loss with loss_type = 'binary'
     loss_cfg = dict(
         type='DiceLoss',

tests/test_models/test_losses/test_lovasz_loss.py

@@ -38,6 +38,36 @@ def test_lovasz_loss():
     labels = (torch.rand(1, 4, 4) * 2).long()
     lovasz_loss(logits, labels, ignore_index=None)
 
+    # test loss with class weights from file
+    import os
+    import tempfile
+    import mmcv
+    import numpy as np
+    tmp_file = tempfile.NamedTemporaryFile()
+
+    mmcv.dump([1.0, 2.0, 3.0], f'{tmp_file.name}.pkl', 'pkl')  # from pkl file
+    loss_cfg = dict(
+        type='LovaszLoss',
+        per_image=True,
+        reduction='mean',
+        class_weight=f'{tmp_file.name}.pkl',
+        loss_weight=1.0)
+    lovasz_loss = build_loss(loss_cfg)
+    lovasz_loss(logits, labels, ignore_index=None)
+
+    np.save(f'{tmp_file.name}.npy', np.array([1.0, 2.0, 3.0]))  # from npy file
+    loss_cfg = dict(
+        type='LovaszLoss',
+        per_image=True,
+        reduction='mean',
+        class_weight=f'{tmp_file.name}.npy',
+        loss_weight=1.0)
+    lovasz_loss = build_loss(loss_cfg)
+    lovasz_loss(logits, labels, ignore_index=None)
+    tmp_file.close()
+    os.remove(f'{tmp_file.name}.pkl')
+    os.remove(f'{tmp_file.name}.npy')
+
     # test lovasz loss with loss_type = 'binary' and per_image = False
     loss_cfg = dict(
         type='LovaszLoss',