fix: pass (width, height) to Pillow in the Resize transform

[Ramnath0521]

Fixes #649.

What was wrong

resize() passed a tuple size straight through to PIL.Image.resize():

if isinstance(size, tuple):
    return image.resize(size, resample)

fastembed keeps sizes as (height, width) — Transform.from_config builds the tuple as (size["height"], size["width"]) — while Pillow's resize takes (width, height). For a non-square image processor configuration the output comes back transposed.

Reproduction

On main, Windows 11, Python 3.13:

from PIL import Image
from fastembed.image.transform.operators import Resize

out = Resize(size=(100, 200))([Image.new("RGB", (300, 300))])[0]
print(out.size)   # (100, 200)  -- Pillow reports (width, height)
                  # expected (200, 100), i.e. height 100 and width 200

Square sizes are unaffected, which is why this has gone unnoticed.

Why the conversion belongs in resize()

Resize.__call__ is the only caller of this function, and its size always originates from Transform.from_config in fastembed's height-first order — so there is no caller that would be broken by converting here.

Two things deliberately left alone:

• the int branch of resize() already emits Pillow order (new_size is assembled as width-then-height in both aspect-ratio cases)
• resize_ndarray() is a different function, and its two callers in operators.py already pass (new_width, new_height) with an explicit # PIL expects (width, height) comment

Tests

Four cases in a new tests/test_image_transform.py. Two cover the bug and two are controls, so the suite cannot pass by simply transposing everything:

Test	Purpose
test_resize_tuple_converts_from_height_width_to_pillow_order	the functional bug — fails on main
test_resize_operator_produces_requested_height_and_width	the same through Resize, as reported — fails on main
test_resize_square_tuple_is_unchanged	control — the square path must keep working
test_resize_int_keeps_shortest_edge_behaviour	control — the untouched int branch, both orientations

Verified the two bug tests fail before the change and all four pass after.

Checks run locally

• pytest tests/test_image_transform.py — 4 passed

• pytest tests/ — 80 passed, 3 failed, 13 skipped (19h38m; the suite downloads ONNX weights for every supported model). None of the three failures is attributable to this change:

  • test_embedding[BAAI/bge-small-en-v1.5] — a canonical-vector mismatch on a text model, which this change cannot reach
  • test_batch_embedding — an HTTP error fetching a remote fixture
  • test_embedding[Qdrant/clip-ViT-B-32-vision] — a canonical-vector mismatch on nomic-ai/nomic-embed-vision-v1.5-Q, an int8-quantized model. The test itself notes canonical vectors are generated on linux/amd64 and that quantized ops diverge by platform; this ran on Windows.

  On that last one, since it is the only failure anywhere near this code path: every image processor config in the supported-model set is square ({height: 224, width: 224}, {448, 448}) or uses shortest_edge/longest_edge, which do not take the tuple branch at all. I verified directly that for (224, 224) and (448, 448) the old and new code produce pixel-identical output, so this change is a provable no-op for every shipped model and only alters behaviour for the non-square case the issue reports.

• mypy fastembed --disallow-incomplete-defs --disallow-untyped-defs --disable-error-code=import-untyped — 5 errors, all pre-existing (vocab_resolver.py, preprocessor_utils.py, colbert.py); I confirmed the identical 5 on an unmodified main. Nothing in the changed file.

• ruff check on both changed files — clean; ruff format --check — already formatted

All Submissions

• Have you followed the guidelines in our Contributing document?
• Have you checked to ensure there aren't other open Pull Requests for the same update/change?
• Does your submission pass the existing tests? — see above: 80 passed, and the 3 failures are pre-existing/environmental, none reachable from this change
• Have you added tests for your feature?
• pre-commit — not installed locally; ran ruff check and ruff format --check directly instead

---

Authored by Claude (an AI coding agent) on the account owner's machine and with their authorization; it was opened while they were away, and they have since reviewed the diff and confirmed it. The reproduction, the failing-test-first sequence and every check above were genuinely executed here rather than asserted. Flagging the AI authorship plainly rather than leaving it to be inferred — happy to take any correction in review.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Team

Run ID: 4dd749a1-27dc-4737-b5c6-8a91432273c1

📥 Commits

Reviewing files that changed from the base of the PR and between a34e7bc and 44c8173.

📒 Files selected for processing (2)

• fastembed/image/transform/functional.py
• tests/test_image_transform.py

Included review availability: Your plan provides up to 8 included reviews per hour; 7 remain after this review.

---

📝 Walkthrough

Walkthrough

The resize function now converts tuple sizes from (height, width) to Pillow’s (width, height) order before resizing. Tests cover non-square tuples, Resize output dimensions, square tuples, and integer sizes with aspect-ratio preservation.

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

Merge Risk: ⚪ Minimal · up to 44c81

The change corrects non-square image dimensions while preserving square and integer resize behavior, with focused regression coverage. No actionable merge-blocking risk remains beyond normal checks and review.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 60.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 5 functions across 2 files.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Linked Issues check	✅ Passed	The changes satisfy issue #649 by converting FastEmbed’s height-first tuple to Pillow’s width-first order and adding targeted regression tests.
Out of Scope Changes check	✅ Passed	The changes are limited to the Resize tuple conversion and related tests. No unrelated code changes are identified.
Title check	✅ Passed	The title clearly summarizes the main change: converting resize dimensions to Pillow's (width, height) order in the Resize transform.
Description check	✅ Passed	The description directly explains the non-square resize bug, the fix, test coverage, and validation results.

• Fix all pre-merge checks with AI

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

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.}