fix: is_empty() returns False for empty tracker arrays

[abritton2002]

Summary

Fixes #2195.

is_empty() incorrectly returned False on genuinely empty Detections objects when tracker_id was an empty numpy array rather than None.

Root cause: The previous implementation compared self against Detections.empty() via __eq__, which uses np.array_equal. When a Detections object with tracker_id set is filtered to zero results, tracker_id becomes np.array([]) (not None). np.array_equal(np.array([]), None) returns False, so the equality check failed and is_empty() returned False.

Fix: Replace the comparison-based approach with a direct length check on self.xyxy, which is the canonical source of truth for detection count:

# before
def is_empty(self) -> bool:
    empty_detections = Detections.empty()
    empty_detections.data = self.data
    empty_detections.metadata = self.metadata
    return bool(self == empty_detections)

# after
def is_empty(self) -> bool:
    return len(self.xyxy) == 0

Test plan

• Added 4 parametrized regression tests in tests/detection/test_core.py::test_is_empty
  • Canonical Detections.empty() → True
  • Non-empty detections without tracker_id → False
  • Filtered-to-empty detections with tracker_id (the regression case) → True
  • One detection remaining after filter → False
• All tests pass: uv run python -m pytest tests/detection/test_core.py::test_is_empty -v

🤖 Generated with Claude Code

[CLAassistant]

All committers have signed the CLA.

[Copilot]

Pull request overview

Fixes sv.Detections.is_empty() incorrectly returning False for filtered-to-empty detections when tracker_id is an empty NumPy array (instead of None), aligning emptiness with the canonical detection count (xyxy length).

Changes:

• Reimplemented Detections.is_empty() to use len(self.xyxy) == 0 rather than equality vs Detections.empty().
• Added parametrized regression tests covering empty/non-empty cases including the tracker_id filtered-to-empty scenario.
• Expanded the is_empty() docstring to describe semantics and provide an example.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
src/supervision/detection/core.py	Updates is_empty() implementation and docstring to define emptiness by xyxy length.
tests/detection/test_core.py	Adds regression coverage for is_empty() including the tracker_id empty-array case.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 78%. Comparing base (6b5eed2) to head (8244808).
⚠️ Report is 1 commits behind head on develop.

Additional details and impacted files

@@           Coverage Diff           @@
##           develop   #2209   +/-   ##
=======================================
- Coverage       78%     78%   -0%     
=======================================
  Files           63      63           
  Lines         7977    7972    -5     
=======================================
- Hits          6253    6248    -5     
  Misses        1724    1724           

🚀 New features to boost your workflow:

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

[Copilot]

Pull request overview

Fixes sv.Detections.is_empty() incorrectly returning False for “filtered-to-empty” detections when optional fields (e.g., tracker_id, mask) are present as empty arrays instead of None, by making emptiness depend solely on detection count (xyxy length).

Changes:

• Simplified Detections.is_empty() to return len(self.xyxy) == 0.
• Updated merge-related expectations in tests/detection/test_core.py now that truly empty detections are correctly recognized and ignored by Detections.merge.
• Added parametrized regression tests for is_empty() covering the tracker_id empty-array case (and a similar mask case).

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
src/supervision/detection/core.py	Replaces equality-based emptiness detection with an xyxy length check and expands the docstring.
tests/detection/test_core.py	Adds regression tests for is_empty() and updates merge test expectations to align with corrected emptiness handling.

[Borda]

is_empty() Robustness Across Detection Types

Detection type	xyxy always populated?	is_empty() correct?
Bounding box	Yes — required field	✓
OBB	Yes — axis-aligned wrapper always set	✓
Segmentation (dense mask)	Yes — computed via mask_to_xyxy()	✓
Segmentation (CompactMask)	Yes — same, never materialized	✓ + O(1)
VLM outputs (polygons, etc.)	Yes — connectors compute xyxy first	✓
Empty (cls.empty())	Yes — (0, 4) shape	✓
KeyPoints	N/A — separate class	⚠ same old bug, unfixed