fix(detection): support float64 and arbitrary dtypes in polygon utilities

[pranaysb]

Before submitting

• Self-reviewed the code
• Updated documentation, follow Google-style
• Added docs entry for autogeneration (if new functions/classes)
• Added/updated tests
• All tests pass locally

Description

Support arbitrary polygon dtypes (such as np.float64, NumPy's default float, as well as int64, int16, uint8) in filter_polygons_by_area and approximate_polygon.

Type of Change

• 🐛 Bug fix (non-breaking change which fixes an issue)

Motivation and Context

cv2.contourArea and cv2.approxPolyDP require inputs of depth CV_32F (np.float32) or CV_32S (np.int32). When users pass standard NumPy polygons (which default to np.float64 or np.int64 on 64-bit systems), OpenCV throws a C++ assertion failure:

cv2.error: OpenCV(4.x.x) ...: error: (-215:Assertion failed) npoints >= 0 && (depth == CV_32F || depth == CV_32S) in function 'contourArea'

In addition, filter_polygons_by_area had a minor typo referencing ares instead of areas.

Changes Made

• In filter_polygons_by_area, pass np.float32 copy to cv2.contourArea if the input polygon is not float32 or int32. Fixed typo ares -> areas.
• In approximate_polygon, ensure OpenCV receives float32/int32, while preserving and returning points with the input array's original dtype.
• Added parameterized unit tests in tests/detection/utils/test_polygons.py validating behavior across np.float64, np.float32, np.int64, np.int32, np.int16, and np.uint8.

Testing

• I have tested this code locally
• I have added unit tests that prove my fix is effective or that my feature works
• All new and existing tests pass (43/43 tests in tests/detection/utils/test_polygons.py)
• ruff check and ruff format --check passed

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 88%. Comparing base (badd869) to head (d72ac3e).

Additional details and impacted files

@@           Coverage Diff           @@
##           develop   #2542   +/-   ##
=======================================
  Coverage       88%     88%           
=======================================
  Files           85      85           
  Lines        12198   12208   +10     
=======================================
+ Hits         10734   10744   +10     
  Misses        1464    1464           

🚀 New features to boost your workflow:

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

[Copilot]

🟡 Changes recommended

Absolute float32 conversion corrupts large-coordinate polygons, and the required changelog entry is missing.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Pull request overview

Adds broader NumPy dtype support to polygon area filtering and approximation.

Changes:

• Converts unsupported OpenCV dtypes while preserving returned dtypes.
• Adds parametrized dtype tests.
• Fixes the ares typo.

Review: Code quality 2/5; testing 3/5; documentation 2/5; risk 3/5. Large-coordinate precision and changelog issues remain.

File summaries

File	Description
src/supervision/detection/utils/polygons.py	Adds dtype conversion for OpenCV polygon operations.
tests/detection/utils/test_polygons.py	Tests supported input and output dtypes.

Review details

Suppressed comments (1)

tests/detection/utils/test_polygons.py:188

• Repository test guidance requires non-primitive parametrized cases such as dtype classes to use pytest.param(..., id="semantic-slug") rather than bare values (.github/CONTRIBUTING.md:342). Use explicit IDs here as well so parametrized case names remain stable and clear.

    [np.float64, np.float32, np.int64, np.int32, np.int16, np.uint8],

• Files reviewed: 2/2 changed files
• Comments generated: 4
• Review effort level: Balanced

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.