Fix Pad.inverse channel-dimension back-crop#8900
Open
aymuos15 wants to merge 2 commits into
Open
Conversation
Pad.inverse computed the channel-dimension back-crop as min(max(padded[0][1], s + 1), len(data)), where s is the front pad. The max(..., s + 1) term forced the crop to be at least front_pad + 1 even when the recorded back pad was smaller (or zero), so inverting a transform with a non-zero channel pad sliced away the wrong number of channels (often emptying the channel dim) and failed to round-trip. Compute the back-crop directly from the recorded back pad instead: e = min(padded[0][1], len(data) - s). Signed-off-by: Soumya Snigdha Kundu <soumya_snigdha.kundu@kcl.ac.uk>
Covers round-tripping a non-zero channel pad (front-only, back-only, symmetric, and combined channel+spatial) through Pad/Pad.inverse using the NumPy backend, which pads the channel dimension. These cases failed before the back-crop fix. Signed-off-by: Soumya Snigdha Kundu <soumya_snigdha.kundu@kcl.ac.uk>
Contributor
📝 WalkthroughWalkthroughThe PR fixes a boundary-condition bug in the Estimated code review effort🎯 2 (Simple) | ⏱️ ~12 minutes 🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@tests/transforms/croppad/test_pad_inverse_channel.py`:
- Around line 32-40: Add a Google-style docstring to the test method
test_inverse_roundtrip_channel_pad describing the test's purpose and its
parameter to_pad (e.g., what padding configuration is being tested and the
expected round-trip behavior), placed immediately below the def line; ensure it
follows Google style with a short description and an Args section for to_pad.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
Run ID: 305f534c-ba43-49b2-9512-d937a43c10e1
📒 Files selected for processing (2)
monai/transforms/croppad/array.pytests/transforms/croppad/test_pad_inverse_channel.py
Comment on lines
+32
to
+40
| def test_inverse_roundtrip_channel_pad(self, to_pad): | ||
| set_track_meta(True) | ||
| img = MetaTensor(torch.arange(3 * 4 * 4).reshape(3, 4, 4).float()) | ||
| pad = Pad(to_pad=to_pad, mode="symmetric") | ||
| padded = pad(img.clone()) | ||
| self.assertEqual(padded.shape[0], img.shape[0] + to_pad[0][0] + to_pad[0][1]) | ||
| inv = pad.inverse(padded) | ||
| self.assertEqual(inv.shape, img.shape) | ||
| self.assertTrue(torch.equal(inv.as_tensor(), img.as_tensor())) |
Contributor
There was a problem hiding this comment.
Add docstring to test method.
Per coding guidelines, all definitions require Google-style docstrings describing parameters and purpose.
📝 Suggested docstring
`@parameterized.expand`(CHANNEL_PAD_CASES)
def test_inverse_roundtrip_channel_pad(self, to_pad):
+ """Test Pad.inverse correctly restores original tensor after channel-dimension padding.
+
+ Args:
+ to_pad: Padding specification [(ch_pad), (h_pad), (w_pad)].
+ """
set_track_meta(True)📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
Suggested change
| def test_inverse_roundtrip_channel_pad(self, to_pad): | |
| set_track_meta(True) | |
| img = MetaTensor(torch.arange(3 * 4 * 4).reshape(3, 4, 4).float()) | |
| pad = Pad(to_pad=to_pad, mode="symmetric") | |
| padded = pad(img.clone()) | |
| self.assertEqual(padded.shape[0], img.shape[0] + to_pad[0][0] + to_pad[0][1]) | |
| inv = pad.inverse(padded) | |
| self.assertEqual(inv.shape, img.shape) | |
| self.assertTrue(torch.equal(inv.as_tensor(), img.as_tensor())) | |
| def test_inverse_roundtrip_channel_pad(self, to_pad): | |
| """Test Pad.inverse correctly restores original tensor after channel-dimension padding. | |
| Args: | |
| to_pad: Padding specification [(ch_pad), (h_pad), (w_pad)]. | |
| """ | |
| set_track_meta(True) | |
| img = MetaTensor(torch.arange(3 * 4 * 4).reshape(3, 4, 4).float()) | |
| pad = Pad(to_pad=to_pad, mode="symmetric") | |
| padded = pad(img.clone()) | |
| self.assertEqual(padded.shape[0], img.shape[0] + to_pad[0][0] + to_pad[0][1]) | |
| inv = pad.inverse(padded) | |
| self.assertEqual(inv.shape, img.shape) | |
| self.assertTrue(torch.equal(inv.as_tensor(), img.as_tensor())) |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@tests/transforms/croppad/test_pad_inverse_channel.py` around lines 32 - 40,
Add a Google-style docstring to the test method
test_inverse_roundtrip_channel_pad describing the test's purpose and its
parameter to_pad (e.g., what padding configuration is being tested and the
expected round-trip behavior), placed immediately below the def line; ensure it
follows Google style with a short description and an Args section for to_pad.
Source: Coding guidelines
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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.
Description
Pad.inversecomputed the channel back-crop asmin(max(padded[0][1], s + 1), len(data)), wheresis the front pad. Themax(..., s + 1)term forced the back-crop to at leastfront_pad + 1even when the recorded back pad was smaller or zero, so inverting aPadwith a non-zero channel pad cropped the wrong number of channels (often emptying the channel dim) and failed to round-trip.For example, inverting a channel pad of
(2, 0)on a 3-channel image produceds = 2,e = min(max(0, 3), 5) = 3, i.e.data[2 : len - 3], an empty slice.The back-crop is now computed directly from the recorded back pad:
e = min(padded[0][1], len(data) - s).Note this only affects the NumPy padding backend (e.g.
mode="symmetric"), which pads the channel dimension; the PyTorch backend ignores channel padding, so the added test usesmode="symmetric"to exercise the channel path.Types of changes
./runtests.sh -f -u --net --coverage../runtests.sh --quick --unittests --disttests.