* VideoInfo.fps returns float instead of int
Truncating the raw CAP_PROP_FPS value with int() causes timing drift for
non-integer frame rates (23.976, 29.97, 59.94). Over a long video this
accumulates into noticeable sync errors — e.g. 23 vs 23.976 drifts ~1s
per minute of footage.
Changes:
- VideoInfo.fps type annotation: int -> float
- from_video_path: int(video.get(CAP_PROP_FPS)) -> float(...)
- ByteTrack.frame_rate type annotation: int -> float (already converts
to int internally via max_time_lost = int(frame_rate / 30.0 * buffer))
- Tests: assert fps is float, add float_fps_video_path fixture at 23.976
* fix: update examples to cast float fps to int where required
* fix: wrap long docstring line in FPSBasedTimer (ruff E501)
* test: remove unused float_fps_video_path fixture
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-authored-by: Jirka Borovec <6035284+Borda@users.noreply.github.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
* make pixel and kernel size dynamic
* fix: zero-area guard and is-not-None check in Blur/PixelateAnnotator
- Skip loop iteration when clip_boxes produces x2<=x1 or y2<=y1 (zero-area ROI) to prevent cv2.error crash in both annotators
- Replace falsy `or` pattern with explicit `is not None` so kernel_size=0 / pixel_size=0 are not silently treated as dynamic
- Replace hardcoded `cv2.mean(roi)[:3]` with ndim-aware fill: scalar for grayscale, channel-matched tuple for colour images; avoids shape mismatch broadcast error on single-channel frames
- test_annotate_bbox_smaller_than_pixel_size_does_not_raise: guards against the OpenCV resize crash from issue #703 when bbox < pixel_size
- test_annotate_grayscale_image_does_not_raise: normal pixelation path on 2-D grayscale frame
- test_annotate_grayscale_image_small_roi_does_not_raise: avg-fill fallback on 2-D grayscale frame
- Add ValueError guard in BlurAnnotator.__init__ and PixelateAnnotator.__init__ for explicit sizes < 1; previously passed straight to cv2 causing ZeroDivisionError or OpenCV assertion failures
- Add parametrized tests for invalid sizes (0, -1, -10) and zero-area bbox skipping for both annotators
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: jirka <6035284+Borda@users.noreply.github.com>
Co-authored-by: Claude Code <noreply@anthropic.com>
* fix: preserve area and iscrowd from detection data in COCO export
* fix: use np.asarray().item() to satisfy mypy in iscrowd/area extraction
* test: shorten test name to fix ruff E501 line-length violation
* test: verify data["area"] overrides bbox area when mask is present
* test: stricter iscrowd type check (bool subclass fix)
* test: stricter iscrowd type check in preserves_iscrowd_from_data
* test: stricter iscrowd type check in iscrowd_is_int_when_mask_provided
* fix: prefer data["iscrowd"] over geometry when mask is present
---------
Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: Borda <6035284+Borda@users.noreply.github.com>
Co-authored-by: Claude Code <noreply@anthropic.com>
When a user's callback accidentally runs inference on the full image instead
of the provided slice, detections get incorrect offsets applied, causing a
repeating grid pattern. Add a validation check in _run_callback that emits a
SupervisionWarnings warning when any detection coordinate exceeds the slice
dimensions or is negative. An instance flag prevents repeated warnings across
many slices.
- Wrap _out_of_slice_bounds_warned check-and-set in threading.Lock to prevent duplicate warnings under ThreadPoolExecutor with thread_workers > 1
- Change stacklevel=2 to stacklevel=1 — under executor.submit the stacklevel=2 frame points into concurrent.futures internals, not user code
- Assert exactly 1 warning fires with thread_workers=4 (validates Lock fix)
- Assert no warning for detection touching but not exceeding slice boundary (pins > vs >= semantics)
- Assert second slicer call does not re-warn (documents once-per-instance semantic)
- Extract warning message into `msg` variable to satisfy E501 line-length limit
---------
Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: Borda <6035284+Borda@users.noreply.github.com>
Co-authored-by: Claude Code <noreply@anthropic.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
This commit adds support for 4-digit and 8-digit hexadecimal color codes
with alpha channel (e.g., #FF00FF80 for 50% transparent magenta).
Changes:
- Extended Color dataclass with optional alpha field (default 255)
- Updated _validate_color_hex to accept lengths 3, 4, 6, 8
- Updated from_hex to parse 4-digit (#RGBA) and 8-digit (#RRGGBBAA) hex codes
- Modified as_hex to return #RRGGBBAA when alpha != 255
- Added as_rgba() and as_bgra() methods
- Added from_rgba_tuple() and from_bgra_tuple() class methods
- Updated __eq__ and __hash__ to include alpha channel
- Added comprehensive unit tests for all RGBA functionality
- Updated docstrings with examples of new alpha channel support
---------
Co-authored-by: jirka <6035284+Borda@users.noreply.github.com>
Co-authored-by: Claude Code <noreply@anthropic.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Swap Ultralytics YOLO references for RFDETRSmall (rfdetr package) and
rfdetr-small (Roboflow Inference) to showcase RF-DETR as the primary
model in the quickstart and connector examples.
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
* fix: prevent single object from appearing in multiple polygon zones
when checking if a detection is inside a polygon zone, the previous implementation
would clip the bounding box to fit within each ROI's dimensions before calculating
anchor points. This caused the same detection to produce different anchor points
for different ROIs, allowing it to be counted as present in multiple zones.
* Add regression test for PolygonZone trigger issue #1987 and remove unused `frame_resolution_wh` attribute
* refactor(polygon_zone): vectorize trigger() and strengthen tests
Replace the O(n×m) Python double-loop in PolygonZone.trigger() with
vectorized NumPy. Semantics are identical: compute a (num_anchors,
num_detections) in_bounds mask, use np.clip solely for safe fancy-index
access, then AND with the polygon mask and reduce with np.all(axis=0).
Also removes the now-unused `from dataclasses import replace` import and
a latent np.all(axis=1) call on a 1D array.
Test improvements:
- Group into TestPolygonZoneInit / TestPolygonZoneTrigger classes
- Replace the trivially-passing regression (sum=0 on both old and new
code) with adjacent zones + straddling detection that gives sum=2 on
the old clip_boxes implementation and sum=1 on the fix
- Rename tests to describe behaviour, not issue numbers
- Add test_out_of_bounds_anchor_excluded and
test_anchor_on_polygon_boundary_included edge cases
* test(polygon_zone): verify current_count updates with expected results during trigger
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: jirka <6035284+Borda@users.noreply.github.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
* refactor docstrings in `src/supervision/detection/tools`
* Refine type annotations and docstrings in `transformers.py` to include `TensorLike` protocol and enhance segmentation handling
* Enhance type safety in `transformers.py` by introducing `_is_tensor_like` with `TypeGuard` and refining segmentation result handling
* Refine `CSVSink` type annotations by introducing `WriterProtocol` for improved type safety
* Remove `_is_tensor_like`, `TensorLike` protocol, and `TypeGuard` usage from `transformers.py`, streamline segmentation handling, and fix typos in `CSVSink` and `JSONSink` docstrings
---------
Co-authored-by: Jirka Borovec <6035284+Borda@users.noreply.github.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
* refactor docstrings in `scr/supervision/detection`
* Enhance docstrings across multiple modules: clarify attributes/args, improve formatting, and update logic for handling sentinel values in metrics calculation.
* Ensure consistent handling of `class_id` as integer across YOLO and Pascal VOC formats, fix NoneType handling in line zone logic, and add test coverage for multiclass annotator with None `class_id`.
* Enforce `class_id` as integer in YOLO export, update line zone class count docstrings, and add test for non-integer `class_id`.
* Apply suggestions from code review
---------
Co-authored-by: Jirka Borovec <6035284+Borda@users.noreply.github.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
* refactor docstrings in draw, classification, and key_points
* Refactor type annotations, logging, and empty output handling across key modules
* Refactor type annotations in `core.py` to include conditional `TYPE_CHECKING` for `torch` imports
* Apply suggestions from code review
---------
Co-authored-by: Jirka Borovec <6035284+Borda@users.noreply.github.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
* refactor(metrics): update MeanAveragePrecisionResult to use float32 precision
This change reduces memory usage and improves computation speed while maintaining acceptable precision for most computer vision tasks.
- Changed mAP_scores, ap_per_class, and iou_thresholds from float64 to float32
- Updated EPS constant to use float32 precision
- Modified COCOEvaluator parameters and computations to use float32
- Added type annotations for variables in filter_segments_by_distance function
- Changed precision and recall calculations to use float32
- Updated average precision computation to handle float32 values
- Modified summarize_predictions to return float32 array
* update type annotation for `keep_labels` to use `npt.NDArray[np.bool_]`
* update EPS constant and mAP calculations for precision consistency
* Apply suggestions from code review
---------
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
DetectionDataset.__init__ now maps class_id to class names using
CLASS_NAME_DATA_FIELD so that LabelAnnotator displays human-readable
labels instead of raw integer IDs.
---------
Co-authored-by: Jirka Borovec <6035284+Borda@users.noreply.github.com>