sigmoid focal loss
class SigmoidFocalLoss(nn.Module):
def __init__(self, ignore_label, gamma=2.0, alpha=0.25,
reduction='mean'):
super(SigmoidFocalLoss, self).__init__()
self.ignore_label = ignore_label
self.gamma = gamma
self.alpha = alpha
self.reduction = reduction
def forward(self, pred, target):
b, h, w = target.size()
pred = pred.view(b, -1, 1)
pred_sigmoid = pred.sigmoid()
target = target.view(b, -1).float()
mask = (target.ne(self.ignore_label)).float()
target = mask * target
onehot = target.view(b, -1, 1)
# TODO: use the pred instead of pred_sigmoid
max_val = (-pred_sigmoid).clamp(min=0)
pos_part = (1 - pred_sigmoid) ** self.gamma * (
pred_sigmoid - pred_sigmoid * onehot)
neg_part = pred_sigmoid ** self.gamma * (max_val + (
(-max_val).exp() + (-pred_sigmoid - max_val).exp()).log())
loss = -(self.alpha * pos_part + (1 - self.alpha) * neg_part).sum(
dim=-1) * mask
if self.reduction == 'mean':
loss = loss.mean()
return loss
ohem
class ProbOhemCrossEntropy2d(nn.Module):
def __init__(self, ignore_label, reduction='mean', thresh=0.6, min_kept=256,
down_ratio=1, use_weight=False):
super(ProbOhemCrossEntropy2d, self).__init__()
self.ignore_label = ignore_label
self.thresh = float(thresh)
self.min_kept = int(min_kept)
self.down_ratio = down_ratio
if use_weight:
weight = torch.FloatTensor(
[1.4297, 1.4805, 1.4363, 3.365, 2.6635, 1.4311, 2.1943, 1.4817,
1.4513, 2.1984, 1.5295, 1.6892, 3.2224, 1.4727, 7.5978, 9.4117,
15.2588, 5.6818, 2.2067])
self.criterion = torch.nn.CrossEntropyLoss(reduction=reduction,
weight=weight,
ignore_index=ignore_label)
else:
self.criterion = torch.nn.CrossEntropyLoss(reduction=reduction,
ignore_index=ignore_label)
def forward(self, pred, target):
b, c, h, w = pred.size()
target = target.view(-1)
valid_mask = target.ne(self.ignore_label)
target = target * valid_mask.long()
num_valid = valid_mask.sum()
prob = F.softmax(pred, dim=1)
prob = (prob.transpose(0, 1)).reshape(c, -1)
if self.min_kept > num_valid:
logger.info('Labels: {}'.format(num_valid))
elif num_valid > 0:
prob = prob.masked_fill_(1 - valid_mask, 1)
mask_prob = prob[
target, torch.arange(len(target), dtype=torch.long)]
threshold = self.thresh
if self.min_kept > 0:
_, index = torch.sort(mask_prob)
threshold_index = index[min(len(index), self.min_kept) - 1]
if mask_prob[threshold_index] > self.thresh:
threshold = mask_prob[threshold_index]
kept_mask = mask_prob.le(threshold)
target = target * kept_mask.long()
valid_mask = valid_mask * kept_mask
# logger.info('Valid Mask: {}'.format(valid_mask.sum()))
target = target.masked_fill_(1 - valid_mask, self.ignore_label)
target = target.view(b, h, w)
return self.criterion(pred, target)
出处:TorchSeg