fix(segmentation): handle empty FOVs + migrate off deprecated skimage min_size= by alxndrkalinin · Pull Request #44 · alxndrkalinin/cubic

alxndrkalinin · 2026-05-29T04:34:12Z

Summary

Two unrelated cleanups around cleanup_segmentation:

Empty FOVs no longer crash. check_labeled_binary previously did assert len(unique_values) > 1, \"Image is constant.\" — so an all-zero label mask (a legitimate "no cells detected" FOV) crashed mid-pipeline with AssertionError, and assert is stripped under python -O anyway. Drop that check (downstream steps are no-ops on constant images) and convert the dtype assert to a proper TypeError. The single-label UserWarning stays.
No more min_size= deprecation noise. skimage.morphology.remove_small_objects deprecated min_size= in 0.26 in favor of max_size= (with inverted threshold semantics — removal target is 0.28). cucim 26.02 still uses min_size=. Added a tiny _remove_small_objects compat helper that bypasses the cubic.skimage proxy for this one call and dispatches per device with the right kwarg: max_size=min_size - 1 for skimage ≥0.26 (preserves the old "keep size ≥ min_size" semantics), min_size= for cucim and older skimage. Both cleanup_segmentation and the public remove_small_objects wrapper route through it.

Test plan

Empty FOV (all-zero label mask) round-trips through cleanup_segmentation with min_obj_size/border_buffer_size and returns unchanged
check_labeled_binary(float32) now raises TypeError (was AssertionError)
cleanup_segmentation on a real label image with one below-threshold and one above-threshold object emits no FutureWarning and removes the small one
GPU path (CuPy input) still works via cucim with min_size=
Existing tests/segmentation/test_clear_border.py passes

🤖 Generated with Claude Code

Summary by Sourcery

Handle empty field-of-view label masks without errors and adapt small-object removal to API changes in skimage and cucim.

Bug Fixes:

Allow constant (e.g., all-zero) labeled images to pass through segmentation checks without raising, while still validating integer dtypes with a proper TypeError.

Enhancements:

Introduce a backend-aware helper for removing small objects that transparently handles skimage's min_size/max_size API change and routes GPU inputs through cucim.
Route segmentation cleanup and the public remove_small_objects helper through the new compatibility function to avoid deprecation warnings and keep behavior consistent across backends.

… min_size= Two unrelated cleanups around `cleanup_segmentation`: 1. `check_labeled_binary` previously `assert`-ed `len(unique_values) > 1`, so passing an all-zero label mask (a legitimate "no cells detected" FOV) crashed mid-pipeline with AssertionError. Asserts are also stripped under `python -O`. Drop that check (downstream steps are no-ops on constant images) and convert the dtype assert to a proper `TypeError`. Keep the single-label warning. 2. `morphology.remove_small_objects(..., min_size=)` emits a FutureWarning in skimage 0.26+ (the kwarg was renamed to `max_size` with inverted threshold semantics, slated for removal in 0.28). cucim 26.02 still uses `min_size=`. Add a tiny `_remove_small_objects` compat helper that bypasses the cubic.skimage proxy for this one call and dispatches per-device with the right kwarg — `max_size=min_size - 1` for skimage ≥0.26, `min_size=` for cucim and older skimage. Preserves "keep size ≥ min_size" semantics across both backends. Verified end-to-end on CPU (empty FOV no longer crashes; bad-dtype now raises TypeError; no FutureWarning leaks through) and GPU (cucim path still routes through `min_size=`). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

sourcery-ai · 2026-05-29T04:34:18Z

Reviewer's Guide

Updates segmentation utilities so empty/constant labeled images no longer assert, and introduces a backend-aware helper around skimage/cucim remove_small_objects to avoid the deprecated min_size argument while preserving existing semantics for both CPU and GPU paths.

Sequence diagram for _remove_small_objects CPU/GPU dispatch

sequenceDiagram
    participant Caller
    participant _remove_small_objects
    participant get_device
    participant cucim_remove_small_objects
    participant skimage_remove_small_objects
    participant inspect

    Caller->>_remove_small_objects: _remove_small_objects(label_img, min_size)
    _remove_small_objects->>get_device: get_device(label_img)
    get_device-->>_remove_small_objects: device
    alt device is GPU
        _remove_small_objects->>cucim_remove_small_objects: remove_small_objects(label_img, min_size=min_size)
        cucim_remove_small_objects-->>_remove_small_objects: label_img
    else device is CPU
        _remove_small_objects->>inspect: inspect.signature(skimage_remove_small_objects)
        inspect-->>_remove_small_objects: signature
        alt max_size in signature.parameters
            _remove_small_objects->>skimage_remove_small_objects: remove_small_objects(label_img, max_size=min_size - 1)
        else
            _remove_small_objects->>skimage_remove_small_objects: remove_small_objects(label_img, min_size=min_size)
        end
        skimage_remove_small_objects-->>_remove_small_objects: label_img
    end
    _remove_small_objects-->>Caller: label_img

File-Level Changes

Change	Details	Files
Relax labeled image validation to accept constant/empty FOVs and raise a real TypeError for non-integer inputs.	Replace the dtype assert in `check_labeled_binary` with an explicit `TypeError` when the image is not integer-typed. Remove the assertion that rejected constant images (e.g., all-zero label masks), allowing empty FOVs to pass silently. Retain the single-label `UserWarning` when exactly two unique values are present in the label image.	`cubic/segmentation/segment_utils.py`
Introduce a compatibility wrapper for `remove_small_objects` to handle skimage 0.26+ `max_size` semantics and cucim's still-supported `min_size`, and route existing callers through it.	Add `_remove_small_objects` helper that dispatches based on device (CPU vs GPU) and inspects the skimage `remove_small_objects` signature to decide between `max_size` and `min_size` arguments, mapping `min_size` to `max_size=min_size-1` when needed. Update `cleanup_segmentation` to use `_remove_small_objects` instead of `morphology.remove_small_objects` directly for `min_obj_size` filtering. Update the public `remove_small_objects` function to call `_remove_small_objects` after running `check_labeled_binary`, returning its result directly.	`cubic/segmentation/segment_utils.py`

Tips and commands

Interacting with Sourcery

Trigger a new review: Comment @sourcery-ai review on the pull request.
Continue discussions: Reply directly to Sourcery's review comments.
Generate a GitHub issue from a review comment: Ask Sourcery to create an
issue from a review comment by replying to it. You can also reply to a
review comment with @sourcery-ai issue to create an issue from it.
Generate a pull request title: Write @sourcery-ai anywhere in the pull
request title to generate a title at any time. You can also comment
@sourcery-ai title on the pull request to (re-)generate the title at any time.
Generate a pull request summary: Write @sourcery-ai summary anywhere in
the pull request body to generate a PR summary at any time exactly where you
want it. You can also comment @sourcery-ai summary on the pull request to
(re-)generate the summary at any time.
Generate reviewer's guide: Comment @sourcery-ai guide on the pull
request to (re-)generate the reviewer's guide at any time.
Resolve all Sourcery comments: Comment @sourcery-ai resolve on the
pull request to resolve all Sourcery comments. Useful if you've already
addressed all the comments and don't want to see them anymore.
Dismiss all Sourcery reviews: Comment @sourcery-ai dismiss on the pull
request to dismiss all existing Sourcery reviews. Especially useful if you
want to start fresh with a new review - don't forget to comment
@sourcery-ai review to trigger a new review!

Customizing Your Experience

Access your dashboard to:

Enable or disable review features such as the Sourcery-generated pull request
summary, the reviewer's guide, and others.
Change the review language.
Add, remove or edit custom review instructions.
Adjust other review settings.

Getting Help

Contact our support team for questions or feedback.
Visit our documentation for detailed guides and information.
Keep in touch with the Sourcery team by following us on X/Twitter, LinkedIn or GitHub.

sourcery-ai

Hey - I've left some high level feedback:

In _remove_small_objects, calling inspect.signature on every invocation adds avoidable overhead in hot paths; consider feature-detecting max_size vs min_size once at import time and reusing that result.
_remove_small_objects translates min_size to max_size=min_size - 1 without handling edge cases like min_size <= 0; it may be safer to explicitly guard or short-circuit for non-positive values to preserve prior behavior and avoid surprising arguments to skimage.
The _remove_small_objects and cleanup_segmentation type hints use np.ndarray while the code explicitly supports CuPy arrays (GPU path), so widening the annotations (e.g., to ArrayLike or a union) would better reflect the accepted input types and avoid confusion for callers and type checkers.

Prompt for AI Agents

Please address the comments from this code review:

## Overall Comments
- In `_remove_small_objects`, calling `inspect.signature` on every invocation adds avoidable overhead in hot paths; consider feature-detecting `max_size` vs `min_size` once at import time and reusing that result.
- `_remove_small_objects` translates `min_size` to `max_size=min_size - 1` without handling edge cases like `min_size <= 0`; it may be safer to explicitly guard or short-circuit for non-positive values to preserve prior behavior and avoid surprising arguments to skimage.
- The `_remove_small_objects` and `cleanup_segmentation` type hints use `np.ndarray` while the code explicitly supports CuPy arrays (GPU path), so widening the annotations (e.g., to `ArrayLike` or a union) would better reflect the accepted input types and avoid confusion for callers and type checkers.

Sourcery is free for open source - if you like our reviews please consider sharing them ✨

_{Help me be more useful! Please click 👍 or 👎 on each comment and I'll use the feedback to improve your reviews.}

Copilot

Pull request overview

This PR updates segmentation cleanup utilities to tolerate empty labeled masks and avoid small-object removal deprecation warnings across CPU/GPU backends.

Changes:

Converts labeled-image dtype validation from assert to TypeError and allows constant/empty FOV masks.
Adds backend-aware _remove_small_objects compatibility helper for skimage/cuCIM API differences.
Routes cleanup and public small-object removal through the new helper.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

…mall_holes helper Two follow-ups from PR review: 1. **Hoist inspect.signature to import time** (Sourcery): the skimage `max_size`-vs-`min_size` feature detection now happens once at module load as `_SKIMAGE_USES_MAX_SIZE`, instead of on every CPU call of `_remove_small_objects`. No per-call inspect cost. 2. **Cover `remove_small_holes` too** (code-review gap): skimage 0.26 deprecated `area_threshold` → `max_size` on `remove_small_holes` with the same pattern (cucim 26.02 still uses `area_threshold`). Add `_remove_small_holes` helper following the same per-device dispatch and wire it into the two existing callsites (`cleanup_segmentation` and `fill_holes_slicer`). Unlike `remove_small_objects`, both old `area_threshold=N` and new `max_size=N` keep identical semantics (no off-by-one). Verified end-to-end on CPU + GPU with FutureWarning set to error. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

…wargs CI's mypy was running against a skimage version that doesn't expose `max_size` on `remove_small_objects` / `remove_small_holes`, so the runtime-feature-detected branches (guarded by `_SKIMAGE_USES_MAX_SIZE`) were flagged as `[call-arg]` errors. Add narrow `# type: ignore[call-arg]` on just those two call sites — the legacy branches still type-check against the older API. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Resets the version after v0.7.0a8/a9/a10 tags were deleted before publish. Covers PRs #42, #43, #44, #45, and #46 on top of v0.7.0a7. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

sourcery-ai Bot reviewed May 29, 2026

View reviewed changes

alxndrkalinin requested a review from Copilot May 29, 2026 04:36

Copilot started reviewing on behalf of alxndrkalinin May 29, 2026 04:36 View session

Copilot AI reviewed May 29, 2026

View reviewed changes

alxndrkalinin and others added 2 commits May 28, 2026 21:44

alxndrkalinin merged commit 0e37ecd into main May 29, 2026
9 checks passed

alxndrkalinin deleted the fix/cleanup-segmentation-empty-fov-and-deprecated-min-size branch May 29, 2026 04:51

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix(segmentation): handle empty FOVs + migrate off deprecated skimage min_size=#44

fix(segmentation): handle empty FOVs + migrate off deprecated skimage min_size=#44
alxndrkalinin merged 3 commits into
mainfrom
fix/cleanup-segmentation-empty-fov-and-deprecated-min-size

alxndrkalinin commented May 29, 2026 •

edited by sourcery-ai Bot

Loading

Uh oh!

sourcery-ai Bot commented May 29, 2026 •

edited

Loading

Interacting with Sourcery

Customizing Your Experience

Getting Help

Uh oh!

sourcery-ai Bot left a comment

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

alxndrkalinin commented May 29, 2026 • edited by sourcery-ai Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Test plan

Summary by Sourcery

Uh oh!

sourcery-ai Bot commented May 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Reviewer's Guide

Sequence diagram for _remove_small_objects CPU/GPU dispatch

File-Level Changes

Interacting with Sourcery

Customizing Your Experience

Getting Help

Uh oh!

sourcery-ai Bot left a comment

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

alxndrkalinin commented May 29, 2026 •

edited by sourcery-ai Bot

Loading

sourcery-ai Bot commented May 29, 2026 •

edited

Loading