Question about MaskFormer: Why the target for dice loss and focal loss are different?

[YTEP-ZHI]

@chhluo Thanks for your hard work by adding maskformer to mmdet. I'm wondering why the target for FocalLoss is 1-mask_targets (as below),

mmdetection/mmdet/models/dense_heads/maskformer_head.py

Lines 409 to 411 in 3b72b12

	# target is (1 - mask_targets) !!!
	loss_mask = self.loss_mask(
	mask_preds, 1 - mask_targets, avg_factor=num_total_masks * h * w)

while the target for DiceLoss is simply mask_targets?

mmdetection/mmdet/models/dense_heads/maskformer_head.py

Lines 398 to 400 in 3b72b12

	# dice loss
	loss_dice = self.loss_dice(
	mask_preds, mask_targets, avg_factor=num_total_masks)

Why doing so? Looking forward to your reply.