[HuaTNA]

Closes #745.

center_crop returns an undersized image when an input dimension needs an odd number of padding pixels: a 3×3 input cropped to 4×4 currently remains 3×3. With resizing disabled, this can leave differently sized images with inconsistent shapes and prevent OnnxImageModel from constructing its input batch.

Round the leading padding up so it offsets the floor-rounded crop origin. Add regression coverage for PIL and NumPy inputs, odd/even padding, mixed crop/pad axes, and pixel alignment against Pillow's crop result. Two preprocessing tests also cover batches with square and rectangular crop configurations.

Validation

• Before the fix: the new center-crop tests produced 14 failures and 6 passes.
• After the fix: pytest -q tests/test_image_transform.py tests/test_common.py tests/test_preprocessor_utils.py tests/test_postprocess.py — 68 passed.
• Ruff lint/format hooks passed; package mypy passed with the CI flags; pyright tests/type_stub.py reported no errors or warnings.
• Tested on Python 3.11.15, NumPy 2.4.6, Pillow 12.3.0, macOS arm64. The full model/GPU suite was not run. New regression tests require no model downloads.

All Submissions

• Followed the contributing guidelines.
• Checked for existing pull requests covering the same change. fix: round Jina CLIP crop offsets to match Jina's reference #731 changes the separate Jina pad2square path.

AI assistance was used for investigation, implementation, and local test execution.

[coderabbitai]

Navigate logical layers of code changes, visualize relationships, and explore their blast radius.

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Advanced

Run ID: 17d33d83-5392-4434-ab74-2934b313d5d2

📥 Commits

Reviewing files that changed from the base of the PR and between 4de4023 and 199fbeb.

📒 Files selected for processing (1)

• tests/test_image_transform.py

Included review availability: This review used your included allowance. Your plan provides up to 8 included reviews per hour; 5 remain after this review.

---

📝 Walkthrough

Walkthrough

center_crop now rounds the top and left padding offsets up when padding an image to the requested crop size. A test checks pixel placement and confirms that processing 3×3 and 4×4 images together produces a batch shaped (2, 3, 4, 4).

Priority: ➖ Normal

Estimated code review effort: 2 (Simple) | ~10 minutes

Change: Bug fix · Severity of issue fixed: Medium

Suggested reviewers: serhiizghama

Merge Risk: ⚪ Minimal · up to 199fb

The padding correction and batch-shape regression test present no identified merge-blocking risk; merge after normal checks.

🚥 Pre-merge checks | ✅ 3 | ❌ 2

❌ Failed checks (2 warnings)

Check name	Status	Explanation	Resolution
Linked Issues check	⚠️ Warning	Issue #745 requires the configured crop dimensions for padding on either or both axes, including mixed crop/pad cases. center_crop rounds leading padding up, and the remaining regression test verifi…	Add regression tests for #745 covering NumPy inputs, even and one-axis padding, mixed crop/pad axes, and rectangular crop batches. Verify output dimensions and pixel alignment for each relevant case.
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 2 functions across 2 files.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (3 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and concisely describes the fix for center cropping with odd padding.
Description check	✅ Passed	The description explains the center-crop issue, the fix, regression coverage, and validation results.
Out of Scope Changes check	✅ Passed	The center_crop change and the remaining batch regression test both address the odd-padding and inconsistent-batch defect in issue #745. The reviewed changes contain no demonstrated unrelated work.

Full details: Linked Issues check

Explanation

Issue #745 requires the configured crop dimensions for padding on either or both axes, including mixed crop/pad cases. center_crop rounds leading padding up, and the remaining regression test verifies a PIL 3×3-to-4×4 batch case. The whole-PR diff does not include the stated coverage for NumPy inputs, even padding, one-axis padding, mixed crop/pad axes, or rectangular batch configurations. Those cases lack automated regression tests.

• Fix all pre-merge checks with AI

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create a new PR

---

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.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[joein]

thanks for addressing this!