Skip to content

Some metrics fail when predictions provide more planes than ground truth #13

Open
true-real-michael opened this issue Jul 29, 2022 · 0 comments

Comments

@true-real-michael
Copy link

true-real-michael commented Jul 29, 2022

"Метрики опираются на predicted плоскости, а если их мало, то получается качетсво высокое и только OSR, USR орет, но он у нас на краях валится так как плохо настроен", -- Д. С. Ярош

Here is an example where Prediction 2 gives better results on all metrics except recall and under_segmented than Prediction 1, but has poorer quality of extraction

Ground truth, Prediction 1, Prediction 2:
1266
1266
1266

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant