Skip to content

Fix masked loss mean #96

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
basardemir wants to merge 1 commit into from
Closed

Fix masked loss mean #96

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
19 changes: 11 additions & 8 deletions src/icon_registration/losses.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -592,16 +592,17 @@ def __call__(self, image_A, image_B, mask_B=None):
I = image_A
J = image_B
assert I.shape == J.shape, "The shape of image I and J should be the same."

if mask_B is not None:
assert mask_B.shape == J.shape, "The shape of mask and J should be the same."

lncc = self.calculate_lncc(I, J)

if mask_B is not None:
mask = mask_B.clone()
mask[mask != 0] = 1
lncc = lncc * mask
return 1 - torch.sum(lncc) / torch.clamp(torch.sum(mask), min=1)

return 1 - torch.mean(lncc)

Expand All @@ -610,17 +611,18 @@ def __call__(self, image_A, image_B, mask_B=None):
I = image_A
J = image_B
assert I.shape == J.shape, "The shape of image I and J should be the same."

if mask_B is not None:
assert mask_B.shape == J.shape, "The shape of mask and J should be the same."

lncc = self.calculate_lncc(I, J)

if mask_B is not None:
mask = mask_B.clone()
mask[mask != 0] = 1
lncc = lncc * mask

return 1 - torch.sum(lncc**2) / torch.clamp(torch.sum(mask), min=1)

return 1 - torch.mean(lncc**2)

class LNCCOnlyInterpolated(SimilarityBase):
Expand Down Expand Up @@ -824,7 +826,8 @@ def __call__(self, image_A, image_B, mask_B=None):

mask = mask.expand(-1, difference.shape[1], -1, -1, -1)
difference = difference * mask

return torch.sum(difference ** 2) / torch.clamp(torch.sum(mask), min=1)

return torch.mean(difference ** 2)

def flips(phi, in_percentage=False):
Expand Down
Loading