fix(metrics): count false positives on empty-GT images

[Borda]

This pull request updates the mean average precision (mAP) metric calculation for object detection, ensuring that predictions on images with no ground-truth objects (background images) are properly penalized as false positives. This change corrects the prior behavior where such predictions did not reduce the mAP score, making the metric more accurate. The update is accompanied by new tests to verify the improved logic.

Metric calculation improvements:

• Updated from_tensors in src/supervision/metrics/detection.py to treat all predictions on ground-truth-empty images as false positives, ensuring these predictions lower the mAP score as expected.

Testing enhancements:

• Added TestMeanAveragePrecisionBackgroundFalsePositives in tests/metrics/test_detection.py with tests verifying that background false positives reduce mAP, and that normal cases without false positives are unaffected.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 85%. Comparing base (f173905) to head (def82b2).

Additional details and impacted files

@@           Coverage Diff           @@
##           develop   #2397   +/-   ##
=======================================
  Coverage       85%     85%           
=======================================
  Files           70      70           
  Lines        10035   10036    +1     
=======================================
+ Hits          8489    8490    +1     
  Misses        1546    1546           

🚀 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 updates the legacy MeanAveragePrecision.from_tensors computation in src/supervision/metrics/detection.py so that predictions made on images with no ground-truth objects (background images) are counted as false positives and reduce the resulting mAP, aligning the metric with expected detection-evaluation behavior.

Changes:

• Updated MeanAveragePrecision.from_tensors to record all predictions on GT-empty images as false positives in the collected per-image stats.
• Added tests ensuring background false positives reduce map50 and that the GT-present path remains numerically stable.

Reviewed changes

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

File	Description
src/supervision/metrics/detection.py	Adjusts from_tensors stats collection to include GT-empty images so predictions are penalized as false positives.
tests/metrics/test_detection.py	Adds regression tests covering background false positives and a pinned GT-present scenario.