[Fix] Fix the bug in binary_cross_entropy (#1527)

* [Fix] Fix the bug in binary_cross_entropy Fix the bug in binary_cross_entropy 'label.max() <= 1' should mask out ignore_index, since the ignore_index often set as 255. * [Fix] Fix the bug in binary_cross_entropy, add comments As the ignore_index often set as 255, so the binary class label check should mask out ignore_index. Co-authored-by: Miao Zheng <[email protected]> * [Fix] Fix the bug in binary_cross_entropy As the ignore_index often set as 255, so the binary class label check should mask out ignore_index. Co-authored-by: Miao Zheng <[email protected]> Co-authored-by: MeowZheng <[email protected]>
open-mmlab · Apr 29, 2022 · 061b5b4 · 061b5b4
1 parent a87b892
commit 061b5b4
Showing 1 changed file with 4 additions and 1 deletion.
diff --git a/mmseg/models/losses/cross_entropy_loss.py b/mmseg/models/losses/cross_entropy_loss.py
@@ -118,7 +118,10 @@ def binary_cross_entropy(pred,
     if pred.size(1) == 1:
         # For binary class segmentation, the shape of pred is
         # [N, 1, H, W] and that of label is [N, H, W].
-        assert label.max() <= 1, \
+        # As the ignore_index often set as 255, so the
+        # binary class label check should mask out
+        # ignore_index
+        assert label[label != ignore_index].max() <= 1, \
             'For pred with shape [N, 1, H, W], its label must have at ' \
             'most 2 classes'
         pred = pred.squeeze()