Fix: resolve major complex review

[Borda]

• DetectionDataset.__init__ lost its isinstance(images, dict) branch in a refactor: dict keys were stored as paths and _images_in_memory stayed empty, so __getitem__, iteration, and merge() on documented dict-form datasets raised "Could not read image from path"
• restore the branch mirroring ClassificationDataset.__init__, including the deprecation warning for dict input
• add 5 regression tests: getitem, len, iteration, equality, and merge round-trip on in-memory datasets

[codecov]

Codecov Report

❌ Patch coverage is 95.52239% with 3 lines in your changes missing coverage. Please review.
✅ Project coverage is 84%. Comparing base (8692148) to head (2bc83ba).

Additional details and impacted files

@@           Coverage Diff           @@
##           develop   #2388   +/-   ##
=======================================
  Coverage       84%     84%           
=======================================
  Files           70      70           
  Lines         9946   10003   +57     
=======================================
+ Hits          8339    8435   +96     
+ Misses        1607    1568   -39     

🚀 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 restores DetectionDataset support for dict-provided in-memory images (including the existing deprecation warning behavior), and also updates mAP evaluation so MeanAveragePrecision can compute IoU based on masks or oriented bounding boxes when metric_target is set accordingly.

Changes:

• Restore DetectionDataset.__init__ handling for images: dict[str, np.ndarray] by keeping images in _images_in_memory and emitting a deprecation warning (mirroring ClassificationDataset).
• Extend MeanAveragePrecision/COCO evaluation to route IoU computation based on metric_target (BOXES vs MASKS vs ORIENTED_BOUNDING_BOXES), and add regression tests for MASKS/OBB behavior.
• Fix ConfusionMatrix.plot() to avoid assigning NaNs into an integer matrix, and update the PolygonZone counting docs example to match current APIs.

Quality (n/5): Code Quality 4/5, Testing 4/5, Docs 3/5

Reviewed changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
src/supervision/dataset/core.py	Restores dict-images initialization branch and deprecation warning for DetectionDataset.
tests/dataset/test_core.py	Adds regression coverage for dict-backed in-memory DetectionDataset behavior.
src/supervision/metrics/mean_average_precision.py	Adds metric-target-aware IoU computation (boxes vs masks vs OBB) in COCO evaluation path.
tests/metrics/test_mean_average_precision.py	Adds MASKS and OBB regression tests for MeanAveragePrecision metric_target routing and errors.
tests/metrics/test_oriented_bounding_box_metrics.py	Updates OBB mAP test expectations/documentation to reflect OBB IoU routing.
src/supervision/metrics/detection.py	Casts confusion matrix to float for plotting to prevent NaN assignment errors on int matrices.
tests/metrics/test_detection.py	Adds a regression test ensuring ConfusionMatrix.plot() returns a figure for int matrices.
docs/how_to/count_in_zone.md	Updates zone-counting tutorial code to use download_assets, ColorPalette.DEFAULT, and current PolygonZone API.