-
Notifications
You must be signed in to change notification settings - Fork 463
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.
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
[FIX] db loss TF and PT also for training with rotated samples #1396
Conversation
Codecov ReportAttention:
Additional details and impacted files@@ Coverage Diff @@
## main #1396 +/- ##
==========================================
- Coverage 95.76% 95.76% -0.01%
==========================================
Files 155 155
Lines 6950 6941 -9
==========================================
- Hits 6656 6647 -9
Misses 294 294
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. |
@odulcy-mindee @charlesmindee Still some work in progress target building needs to be checked again (especially for the shrunken masks) and the loss comp |
balanced_bce_loss = torch.zeros(1, device=out_map.device) | ||
dice_loss = torch.zeros(1, device=out_map.device) | ||
l1_loss = torch.zeros(1, device=out_map.device) | ||
|
||
# TODO: Still in progress @Oliver @Charles |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The loss comp is mostly borrowed from mmocr to see wht's going wrong (and because of missing time xD)
904f038
to
ed1457a
Compare
ed1457a
to
9ec4ce5
Compare
@odulcy-mindee I think that's also fine now :) |
doctr/models/detection/differentiable_binarization/tensorflow.py
Outdated
Show resolved
Hide resolved
doctr/models/detection/differentiable_binarization/tensorflow.py
Outdated
Show resolved
Hide resolved
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks @felixdittrich92 ! 🚀
This PR:
Any feedback is welcome 🤗