An anchor landing exactly on a polygon corner is counted as inside.

PixelateAnnotator must not crash when the bbox is smaller than pixel_size. Regression test for https://github.com/roboflow/supervision/issues

Empty call after heat accumulated must not trigger RuntimeWarning.

Verify that VertexAnnotator handles empty keypoints without modifying the scene. Scenario: Annotating a scene with no key points det

Verify that EdgeAnnotator handles empty keypoints without modifying the scene. Scenario: Annotating a scene with no key points for e

Scenario: Annotating a scene with no keypoints. Expected: Original scene is returned untouched.

PixelateAnnotator must work on single-channel (grayscale) images. The small-ROI avg-fill branch previously sliced cv2.mean()[:3] into a

Grayscale image with bbox smaller than pixel_size uses scalar avg fill. Exercises the ndim-aware branch added to the small-ROI fallback.

Scenario: Covariance data does not match keypoint dimensions. Expected: Clear shape validation error.

Scenario: Annotating non-empty keypoints without covariance data. Expected: Clear error explaining the expected data field.

Verify returning unmodified scene when no known skeleton matches. Scenario: Key points provided don't match any known or provided sk

Empty call must not poison accumulated heat.

Test that uint8 and bool masks produce identical overlays.

Test that uint8 and bool masks produce identical halos.

Test that uint8 and bool masks produce identical overlays.

Test that an all-False mask leaves the scene unchanged, not corrupted.

Scenario: Annotating keypoints with per-point covariance matrices. Expected: Scene is modified with filled ellipses at keypoint locat

Verify that VertexAnnotator respects custom color and radius settings. Scenario: Annotating a scene with user-specified color and ra

Verify that EdgeAnnotator respects custom-defined skeleton structures. Scenario: Annotating a scene with a custom-defined skeleton s

Verify that VertexAnnotator correctly draws keypoints with default parameters. Scenario: Annotating a scene using default vertex par

Verify correctly draw skeleton edges with default parameters. Scenario: Annotating a scene with default skeleton (e.g., COCO).

Verify that annotation with multiple detections draws all bounding boxes. Scenario: Annotating an image with multiple bounding boxes

Verify that annotation with no detections does not change the image. Scenario: Annotating an image with an empty set of detections.

Test that annotate method returns unmodified image when no detections

Test that annotate method returns unmodified image when no detections

Test that annotate method returns unmodified image when no detections

Test that annotate method returns unmodified image when no detections

Test that annotate method returns unmodified image when no detections

Test that annotate method returns unmodified image when no detections

Test that annotate method returns unmodified image when no detections

Test that annotate method returns unmodified image when no detections

Test that annotate method returns unmodified image when no detections

Test that annotate method returns unmodified image when no detections

Test that annotate method returns unmodified image when no detections

Test that annotate method returns unmodified image when no detections

Test that annotate method returns unmodified image when no detections

Test that annotate method returns unmodified image when no detections

Test that annotate method returns unmodified image when no detections

Test that annotate method returns unmodified image when no detections

Test that annotate method returns unmodified image when no detections

Test that annotate method returns unmodified image when no detections

Test that annotate method returns unmodified image when no detections

Empty detections must not trigger a divide-by-zero RuntimeWarning.

Verify that annotation respects custom NumPy color lookup array. Scenario: Providing a custom NumPy array for color lookup instead o

Verify that annotation with a single detection draws a bounding box. Scenario: Annotating an image with a single bounding box.

Test that annotate method correctly draws a single color box

Single detection must produce visible heat — result differs from input.

Test that annotate method correctly draws a single ellipse

Test that annotate method correctly draws box corners

Test that annotate method correctly draws a circle

Test that annotate method correctly draws a dot

Test that annotate method correctly draws a label

Test that annotate method correctly draws a rich label

Test that annotate method correctly blurs a region

Test that annotate method correctly pixelates a region

Test that annotate method correctly draws a triangle

Test that annotate method correctly draws a round box

Test that annotate method correctly draws a percentage bar

Test that annotate method correctly draws a crop

Test that annotate method correctly draws background overlay

Test that annotate method correctly compares two detections

Test that annotate method correctly draws a single mask

Test that annotate method correctly draws a single polygon from mask

Test that annotate method correctly draws a single halo

Test that annotate method returns unmodified image when no masks

Test that annotate method returns unmodified image when no masks

Test that annotate method returns unmodified image when no masks

Test that annotate method returns unmodified image when no OBB data

Zero-area bounding boxes must be silently skipped, not crash.

Zero-area bounding boxes must be silently skipped, not crash.

Assert no annotated_frame assignment wraps the call in a tuple. Guards against the paren-comma typo: ``annotated_frame = (call(),)`` which cr

CompactMask detection straddling image edge paints via NumPy clip.

CompactMask detections annotate identically to dense bool masks.

Assert no tab group exceeds MkDocs Material's 20-tab rendering limit.

Test area calculation and size-specific mAP functionality.

_rle_area must equal np.sum on the original boolean array.

Test that area from data field is preserved (COCO case).

Area dtype matches the documented per-branch contract.

Regression for #768: exporting train/valid/test splits with the returned ids fed forward yields globally unique image and annotation ids.

When only writing images, the starting ids round-trip unchanged so chaining still works in the images-only branch.

Test different averaging methods

Test different averaging methods

Test different averaging methods

_base48_decode returns raw delta-encoded integers from base-48 string.

_base48_decode raises ValueError on truncated continuation sequences.

_base48_encode converts raw delta integers to base-48 string.

_base48_decode(_base48_encode(v)) == v for any valid delta list.

Test that batch updates with perfect detections get 1.0 mAP

Python list[bool] should behave like boolean masking.