Fix extreme memory usage when loading OBB datasets

[Copilot]

OBB annotation lines have 9 values (class_id x1 y1 x2 y2 x3 y3 x4 y4), which caused _with_seg_mask to return True (it checks for > 5 values). This incorrectly set with_masks=True for OBB data, generating an (N, W, H) boolean mask array per image — tens of GBs for high-resolution datasets.

Changes

• src/supervision/dataset/formats/yolo.py: Guard with_masks with not is_obb in load_yolo_annotations, so OBB annotations never trigger mask generation:

  # Before
  with_masks = force_masks or _with_seg_mask(lines=lines)
  # After
  with_masks = not is_obb and (force_masks or _with_seg_mask(lines=lines))

• tests/dataset/formats/test_yolo.py: Added test_load_yolo_annotations_obb_does_not_generate_masks — creates a minimal OBB dataset on disk, loads it with is_obb=True, and asserts detection.mask is None.

---

🔒 GitHub Advanced Security automatically protects Copilot coding agent pull requests. You can protect all pull requests by enabling Advanced Security for your repositories. Learn more about Advanced Security.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 77%. Comparing base (85ba8be) to head (0a22e09).
⚠️ Report is 1 commits behind head on develop.

Additional details and impacted files

@@           Coverage Diff           @@
##           develop   #2187   +/-   ##
=======================================
  Coverage       77%     77%           
=======================================
  Files           62      62           
  Lines         7637    7637           
=======================================
+ Hits          5903    5908    +5     
+ Misses        1734    1729    -5     

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Copilot]

Pull request overview

This PR prevents extreme memory usage when loading YOLO OBB datasets by ensuring OBB annotation lines (which have >5 tokens) do not trigger segmentation-mask generation in load_yolo_annotations.

Changes:

• Disable with_masks when is_obb=True to avoid allocating per-image (N, H, W) boolean mask arrays for OBB data.
• Add regression tests confirming OBB loading never produces masks (even with force_masks=True) and that non-OBB segmentation still produces masks.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
src/supervision/dataset/formats/yolo.py	Prevents OBB datasets from enabling mask generation based on token count.
tests/dataset/formats/test_yolo.py	Adds coverage to prevent the OBB mask-memory regression and verify segmentation behavior remains intact.