Skip to content

Correct functional.py _compute_metric #659

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

Merged
merged 1 commit into from
Dec 14, 2022
Merged

Correct functional.py _compute_metric #659

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
11 changes: 10 additions & 1 deletion segmentation_models_pytorch/metrics/functional.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -260,14 +260,23 @@ def _compute_metric(
tn = tn.sum()
score = metric_fn(tp, fp, fn, tn, **metric_kwargs)

elif reduction == "macro" or reduction == "weighted":
elif reduction == "macro" :
tp = tp.sum(0)
fp = fp.sum(0)
fn = fn.sum(0)
tn = tn.sum(0)
score = metric_fn(tp, fp, fn, tn, **metric_kwargs)
score = _handle_zero_division(score, zero_division)
score = (score * class_weights).mean()

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

What is the difference with "macro"?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

sum() versus mean() as far as I can see

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@JulienMaille Exactly.
sum() instead of mean() when reduction=="weighted", and class_weights are provided and not set to 1.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

then, I guess we have to ensure that sum of class weights is equal to 1, or at least warn it that is not true

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

line 254 normalizes the provided weights, so it has to sum to one:
class_weights = class_weights / class_weights.sum()
what do you think ?

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah, that's right, thanks

elif reduction == "weighted":
tp = tp.sum(0)
fp = fp.sum(0)
fn = fn.sum(0)
tn = tn.sum(0)
score = metric_fn(tp, fp, fn, tn, **metric_kwargs)
score = _handle_zero_division(score, zero_division)
score = (score * class_weights).sum()

elif reduction == "micro-imagewise":
tp = tp.sum(1)
Expand Down