Add data-aware anchor generator #1251
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
wonjuleee
force-pushed
the
aad_anchor_gen
branch
from
435cc1c
to
7706d68
Compare
wonjuleee
force-pushed
the
aad_anchor_gen
branch
from
9336706
to
9db1680
Compare
sooahleex
reviewed
Jan 30, 2024
Comment on lines
+220
to
+230
| scale_samples, ratio_samples = [], [] | ||
| for target in targets[gt_max_overlaps > 0.1]: | ||
| h = target[3] - target[1] | ||
| w = target[2] - target[0] | ||
| r = h / w | ||
| affine_h = base_size * torch.sqrt(r) | ||
| affine_w = base_size / torch.sqrt(r) | ||
| s = max(h / affine_h, w / affine_w) | ||
|
|
||
| scale_samples.append(s) | ||
| ratio_samples.append(r) |
There was a problem hiding this comment.
Suggested change
| scale_samples, ratio_samples = [], [] | |
| for target in targets[gt_max_overlaps > 0.1]: | |
| h = target[3] - target[1] | |
| w = target[2] - target[0] | |
| r = h / w | |
| affine_h = base_size * torch.sqrt(r) | |
| affine_w = base_size / torch.sqrt(r) | |
| s = max(h / affine_h, w / affine_w) | |
| scale_samples.append(s) | |
| ratio_samples.append(r) | |
| scale_samples = [max(h / (base_size * torch.sqrt(r)), w / (base_size / torch.sqrt(r))) for h, w, r in zip(targets[:, 3] - targets[:, 1], targets[:, 2] - targets[:, 0], targets[:, 3] / targets[:, 2])] | |
| ratio_samples = [h / w for h, w in zip(targets[:, 3] - targets[:, 1], targets[:, 2] - targets[:, 0])] |
I think there are some repeated computation.
Comment on lines
352
to
353
| scales = torch.Tensor(scales).to(self.device).detach().requires_grad_(True) | ||
| ratios = torch.Tensor(ratios).to(self.device).detach().requires_grad_(True) |
There was a problem hiding this comment.
Suggested change
| scales = torch.Tensor(scales).to(self.device).detach().requires_grad_(True) | |
| ratios = torch.Tensor(ratios).to(self.device).detach().requires_grad_(True) | |
| scales = scales.detach().requires_grad_(True) | |
| ratios = ratios.detach().requires_grad_(True) |
There was a problem hiding this comment.
I'm sure there's a way to pre-check for errors that isn't try except, but I can't think of it. If there's a better structure, I'd like to use it instead of try except.
Codecov ReportAttention:
Additional details and impacted files@@ Coverage Diff @@
## develop #1251 +/- ##
===========================================
+ Coverage 80.53% 80.57% +0.04%
===========================================
Files 270 271 +1
Lines 30248 30441 +193
Branches 5907 5930 +23
===========================================
+ Hits 24360 24528 +168
- Misses 4507 4525 +18
- Partials 1381 1388 +7
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. |
sooahleex
approved these changes
Jan 30, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary
How to test
Checklist
License
Feel free to contact the maintainers if that's a concern.