Merge pull request #1258 from roboflow/develop

`supervision-0.21.0` release

.pre-commit-config.yaml

@@ -45,7 +45,7 @@ repos:
 
 
   -   repo: https://github.com/astral-sh/ruff-pre-commit
-      rev: v0.4.3
+      rev: v0.4.7
       hooks:
       -   id: ruff
           args: [--fix, --exit-non-zero-on-fix]

docs/changelog.md

@@ -1,3 +1,83 @@
+### 0.21.0 <small>Jun 5, 2024</small>
+
+- Added [#500](https://github.com/roboflow/supervision/pull/500): [`sv.Detections.with_nmm`](https://supervision.roboflow.com/develop/detection/core/#supervision.detection.core.Detections.with_nmm) to perform non-maximum merging on the current set of object detections.
+
+- Added [#1221](https://github.com/roboflow/supervision/pull/1221): [`sv.Detections.from_lmm`](https://supervision.roboflow.com/develop/detection/core/#supervision.detection.core.Detections.from_lmm) allowing to parse Large Multimodal Model (LMM) text result into [`sv.Detections`](https://supervision.roboflow.com/develop/detection/core/) object. For now `from_lmm` supports only [PaliGemma](https://colab.research.google.com/github/roboflow-ai/notebooks/blob/main/notebooks/how-to-finetune-paligemma-on-detection-dataset.ipynb) result parsing.
+
+```python
+import supervision as sv
+
+paligemma_result = "<loc0256><loc0256><loc0768><loc0768> cat"
+detections = sv.Detections.from_lmm(
+    sv.LMM.PALIGEMMA,
+    paligemma_result,
+    resolution_wh=(1000, 1000),
+    classes=['cat', 'dog']
+)
+detections.xyxy
+# array([[250., 250., 750., 750.]])
+
+detections.class_id
+# array([0])
+```
+
+- Added [#1236](https://github.com/roboflow/supervision/pull/1236): [`sv.VertexLabelAnnotator`](https://supervision.roboflow.com/develop/keypoint/annotators/#supervision.keypoint.annotators.EdgeAnnotator.annotate) allowing to annotate every vertex of a keypoint skeleton with custom text and color.
+
+```python
+import supervision as sv
+
+image = ...
+key_points = sv.KeyPoints(...)
+
+edge_annotator = sv.EdgeAnnotator(
+    color=sv.Color.GREEN,
+    thickness=5
+)
+annotated_frame = edge_annotator.annotate(
+    scene=image.copy(),
+    key_points=key_points
+)
+```
+
+- Added [#1147](https://github.com/roboflow/supervision/pull/1147): [`sv.KeyPoints.from_inference`](https://supervision.roboflow.com/develop/keypoint/core/#supervision.keypoint.core.KeyPoints.from_inference) allowing to create [`sv.KeyPoints`](https://supervision.roboflow.com/develop/keypoint/core/#supervision.keypoint.core.KeyPoints) from [Inference](https://github.com/roboflow/inference) result.
+
+- Added [#1138](https://github.com/roboflow/supervision/pull/1138): [`sv.KeyPoints.from_yolo_nas`](https://supervision.roboflow.com/develop/keypoint/core/#supervision.keypoint.core.KeyPoints.from_yolo_nas) allowing to create [`sv.KeyPoints`](https://supervision.roboflow.com/develop/keypoint/core/#supervision.keypoint.core.KeyPoints) from [YOLO-NAS](https://github.com/Deci-AI/super-gradients/blob/master/YOLONAS.md) result.
+
+- Added [#1163](https://github.com/roboflow/supervision/pull/1163): [`sv.mask_to_rle`](https://supervision.roboflow.com/develop/datasets/utils/#supervision.dataset.utils.rle_to_mask) and [`sv.rle_to_mask`](https://supervision.roboflow.com/develop/datasets/utils/#supervision.dataset.utils.rle_to_mask) allowing for easy conversion between mask and rle formats.
+
+- Changed [#1236](https://github.com/roboflow/supervision/pull/1236): [`sv.InferenceSlicer`](https://supervision.roboflow.com/develop/detection/tools/inference_slicer/) allowing to select overlap filtering strategy (`NONE`, `NON_MAX_SUPPRESSION` and `NON_MAX_MERGE`).
+
+- Changed [#1178](https://github.com/roboflow/supervision/pull/1178): [`sv.InferenceSlicer`](https://supervision.roboflow.com/develop/detection/tools/inference_slicer/) adding instance segmentation model support.
+
+```python
+import cv2
+import numpy as np
+import supervision as sv
+from inference import get_model
+
+model = get_model(model_id="yolov8x-seg-640")
+image = cv2.imread(<SOURCE_IMAGE_PATH>)
+
+def callback(image_slice: np.ndarray) -> sv.Detections:
+    results = model.infer(image_slice)[0]
+    return sv.Detections.from_inference(results)
+
+slicer = sv.InferenceSlicer(callback = callback)
+detections = slicer(image)
+
+mask_annotator = sv.MaskAnnotator()
+label_annotator = sv.LabelAnnotator()
+
+annotated_image = mask_annotator.annotate(
+    scene=image, detections=detections)
+annotated_image = label_annotator.annotate(
+    scene=annotated_image, detections=detections)
+```
+
+- Changed [#1228](https://github.com/roboflow/supervision/pull/1228): [`sv.LineZone`](https://supervision.roboflow.com/develop/detection/tools/line_zone/) making it 10-20 times faster, depending on the use case.
+
+- Changed [#1163](https://github.com/roboflow/supervision/pull/1163): [`sv.DetectionDataset.from_coco`](https://supervision.roboflow.com/develop/datasets/core/#supervision.dataset.core.DetectionDataset.from_coco) and [`sv.DetectionDataset.as_coco`](https://supervision.roboflow.com/develop/datasets/core/#supervision.dataset.core.DetectionDataset.as_coco) adding support for run-length encoding (RLE) mask format.
+
 ### 0.20.0 <small>April 24, 2024</small>
 
 - Added [#1128](https://github.com/roboflow/supervision/pull/1128): [`sv.KeyPoints`](/0.20.0/keypoint/core/#supervision.keypoint.core.KeyPoints) to provide initial support for pose estimation and broader keypoint detection models.

docs/cookbooks.md

@@ -1,7 +1,6 @@
 ---
 template: cookbooks.html
 comments: true
-status: new
 hide:
   - navigation
   - toc

docs/datasets/core.md

@@ -1,5 +1,6 @@
 ---
 comments: true
+status: new
 ---
 
 # Datasets

docs/datasets/utils.md

@@ -0,0 +1,18 @@
+---
+comments: true
+status: new
+---
+
+# Datasets Utils
+
+<div class="md-typeset">
+  <h2><a href="#supervision.dataset.utils.rle_to_mask">rle_to_mask</a></h2>
+</div>
+
+:::supervision.dataset.utils.rle_to_mask
+
+<div class="md-typeset">
+  <h2><a href="#supervision.dataset.utils.mask_to_rle">mask_to_rle</a></h2>
+</div>
+
+:::supervision.dataset.utils.mask_to_rle

docs/detection/annotators.md

@@ -1,6 +1,5 @@
 ---
 comments: true
-status: new
 ---
 
 # Annotators
@@ -285,6 +284,37 @@ status: new
 
     </div>
 
+=== "RichLabel"
+
+    ```python
+    import supervision as sv
+
+    image = ...
+    detections = sv.Detections(...)
+
+    labels = [
+        f"{class_name} {confidence:.2f}"
+        for class_name, confidence
+        in zip(detections['class_name'], detections.confidence)
+    ]
+
+    rich_label_annotator = sv.RichLabelAnnotator(
+        font_path=".../font.ttf",
+        text_position=sv.Position.CENTER
+    )
+    annotated_frame = label_annotator.annotate(
+        scene=image.copy(),
+        detections=detections,
+        labels=labels
+    )
+    ```
+
+    <div class="result" markdown>
+
+    ![label-annotator-example](https://media.roboflow.com/supervision-annotator-examples/label-annotator-example-purple.png){ align=center width="800" }
+
+    </div>
+
 === "Crop"
 
     ```python
@@ -492,6 +522,12 @@ status: new
 
 :::supervision.annotators.core.LabelAnnotator
 
+<div class="md-typeset">
+    <h2><a href="#supervision.annotators.core.RichLabelAnnotator">RichLabelAnnotator</a></h2>
+</div>
+
+:::supervision.annotators.core.RichLabelAnnotator
+
 <div class="md-typeset">
     <h2><a href="#supervision.annotators.core.BlurAnnotator">BlurAnnotator</a></h2>
 </div>

docs/detection/double_detection_filter.md

@@ -0,0 +1,30 @@
+---
+comments: true
+status: new
+---
+
+# Double Detection Filter
+
+<div class="md-typeset">
+  <h2><a href="#supervision.detection.overlap_filter.OverlapFilter">OverlapFilter</a></h2>
+</div>
+
+:::supervision.detection.overlap_filter.OverlapFilter
+
+<div class="md-typeset">
+  <h2><a href="#supervision.detection.overlap_filter.box_non_max_suppression">box_non_max_suppression</a></h2>
+</div>
+
+:::supervision.detection.overlap_filter.box_non_max_suppression
+
+<div class="md-typeset">
+  <h2><a href="#supervision.detection.overlap_filter.mask_non_max_suppression">mask_non_max_suppression</a></h2>
+</div>
+
+:::supervision.detection.overlap_filter.mask_non_max_suppression
+
+<div class="md-typeset">
+  <h2><a href="#supervision.detection.overlap_filter.box_non_max_merge">box_non_max_merge</a></h2>
+</div>
+
+:::supervision.detection.overlap_filter.box_non_max_merge

docs/detection/tools/inference_slicer.md

@@ -1,5 +1,6 @@
 ---
 comments: true
+status: new
 ---
 
 # InferenceSlicer

docs/detection/tools/line_zone.md

@@ -1,5 +1,6 @@
 ---
 comments: true
+status: new
 ---
 
 <div class="md-typeset">

docs/detection/tools/save_detections.md

@@ -1,6 +1,5 @@
 ---
 comments: true
-status: new
 ---
 
 # Save Detections

docs/detection/utils.md

@@ -1,6 +1,5 @@
 ---
 comments: true
-status: new
 ---
 
 # Detection Utils
@@ -17,18 +16,6 @@ status: new
 
 :::supervision.detection.utils.mask_iou_batch
 
-<div class="md-typeset">
-  <h2><a href="#supervision.detection.utils.box_non_max_suppression">box_non_max_suppression</a></h2>
-</div>
-
-:::supervision.detection.utils.box_non_max_suppression
-
-<div class="md-typeset">
-  <h2><a href="#supervision.detection.utils.mask_non_max_suppression">mask_non_max_suppression</a></h2>
-</div>
-
-:::supervision.detection.utils.mask_non_max_suppression
-
 <div class="md-typeset">
   <h2><a href="#supervision.detection.utils.polygon_to_mask">polygon_to_mask</a></h2>
 </div>
@@ -88,3 +75,15 @@ status: new
 </div>
 
 :::supervision.detection.utils.pad_boxes
+
+<div class="md-typeset">
+  <h2><a href="#supervision.detection.utils.contains_holes">contains_holes</a></h2>
+</div>
+
+:::supervision.detection.utils.contains_holes
+
+<div class="md-typeset">
+  <h2><a href="#supervision.detection.utils.contains_multiple_segments">contains_multiple_segments</a></h2>
+</div>
+
+:::supervision.detection.utils.contains_multiple_segments

docs/how_to/detect_and_annotate.md

@@ -1,6 +1,5 @@
 ---
 comments: true
-status: new
 ---
 
 # Detect and Annotate

docs/how_to/detect_small_objects.md

@@ -282,7 +282,7 @@ objects within each, and aggregating the results.
 
     def callback(image_slice: np.ndarray) -> sv.Detections:
         results = model.infer(image_slice)[0]
-        detections = sv.Detections.from_inference(results)
+        return sv.Detections.from_inference(results)
 
     slicer = sv.InferenceSlicer(callback = callback)
     detections = slicer(image)

docs/how_to/save_detections.md

@@ -1,6 +1,5 @@
 ---
 comments: true
-status: new
 ---
 
 # Save Detections

docs/trackers.md

@@ -1,6 +1,5 @@
 ---
 comments: true
-status: new
 ---
 
 # ByteTrack

docs/utils/draw.md

@@ -41,7 +41,7 @@ comments: true
 :::supervision.draw.utils.draw_image
 
 <div class="md-typeset">
-    <h2><a href="#supervision.draw.utils.calculate_optimal_font_scale">calculate_optimal_font_scale</a></h2>
+    <h2><a href="#supervision.draw.utils.calculate_optimal_text_scale">calculate_optimal_text_scale</a></h2>
 </div>
 
 :::supervision.draw.utils.calculate_optimal_text_scale

docs/utils/image.md

@@ -1,6 +1,5 @@
 ---
 comments: true
-status: new
 ---
 
 # Image Utils
@@ -12,7 +11,7 @@ status: new
 :::supervision.utils.image.crop_image
 
 <div class="md-typeset">
-    <h2><a href="#supervision.utils.image.scale_image">crop_image</a></h2>
+    <h2><a href="#supervision.utils.image.scale_image">scale_image</a></h2>
 </div>
 
 :::supervision.utils.image.scale_image

docs/utils/iterables.md

@@ -1,6 +1,5 @@
 ---
 comments: true
-status: new
 ---
 
 # Iterables Utils

mkdocs.yml

@@ -35,19 +35,20 @@ extra_css:
 
 
 nav:
-  - Home: index.md
-  - How to:
+  - Supervision: index.md
+  - Learn:
       - Detect and Annotate: how_to/detect_and_annotate.md
       - Save Detections: how_to/save_detections.md
       - Filter Detections: how_to/filter_detections.md
       - Detect Small Objects: how_to/detect_small_objects.md
-      - Track Objects: how_to/track_objects.md
+      - Track Objects on Video: how_to/track_objects.md
 
-  - API:
+  - Reference - Code API:
     - Detection and Segmentation:
         - Core: detection/core.md
         - Annotators: detection/annotators.md
         - Metrics: detection/metrics.md
+        - Double Detection Filter: detection/double_detection_filter.md
         - Utils: detection/utils.md
     - Keypoint Detection:
         - Core: keypoint/core.md
@@ -61,7 +62,9 @@ nav:
         - Detection Smoother: detection/tools/smoother.md
         - Save Detections: detection/tools/save_detections.md
     - Trackers: trackers.md
-    - Datasets: datasets.md
+    - Datasets:
+        - Core: datasets/core.md
+        - Utils: datasets/utils.md
     - Utils:
         - Video: utils/video.md
         - Image: utils/image.md
@@ -76,7 +79,7 @@ nav:
     - Contributing: contributing.md
     - Code of Conduct: code_of_conduct.md
     - License: license.md
-  - Changelog:
+  - Release Notes:
         - Changelog: changelog.md
         - Deprecated: deprecated.md
 

poetry.lock

@@ -1253,21 +1253,21 @@ test-extra = ["curio", "matplotlib (!=3.2.0)", "nbformat", "numpy (>=1.21)", "pa
 
 [[package]]
 name = "ipywidgets"
-version = "8.1.2"
+version = "8.1.3"
 description = "Jupyter interactive widgets"
 optional = false
 python-versions = ">=3.7"
 files = [
-    {file = "ipywidgets-8.1.2-py3-none-any.whl", hash = "sha256:bbe43850d79fb5e906b14801d6c01402857996864d1e5b6fa62dd2ee35559f60"},
-    {file = "ipywidgets-8.1.2.tar.gz", hash = "sha256:d0b9b41e49bae926a866e613a39b0f0097745d2b9f1f3dd406641b4a57ec42c9"},
+    {file = "ipywidgets-8.1.3-py3-none-any.whl", hash = "sha256:efafd18f7a142248f7cb0ba890a68b96abd4d6e88ddbda483c9130d12667eaf2"},
+    {file = "ipywidgets-8.1.3.tar.gz", hash = "sha256:f5f9eeaae082b1823ce9eac2575272952f40d748893972956dc09700a6392d9c"},
 ]
 
 [package.dependencies]
 comm = ">=0.1.3"
 ipython = ">=6.1.0"
-jupyterlab-widgets = ">=3.0.10,<3.1.0"
+jupyterlab-widgets = ">=3.0.11,<3.1.0"
 traitlets = ">=4.3.1"
-widgetsnbextension = ">=4.0.10,<4.1.0"
+widgetsnbextension = ">=4.0.11,<4.1.0"
 
 [package.extras]
 test = ["ipykernel", "jsonschema", "pytest (>=3.6.0)", "pytest-cov", "pytz"]
@@ -1566,13 +1566,13 @@ test = ["jupyter-server (>=2.0.0)", "pytest (>=7.0)", "pytest-jupyter[server] (>
 
 [[package]]
 name = "jupyterlab"
-version = "4.1.2"
+version = "4.2.0"
 description = "JupyterLab computational environment"
 optional = false
 python-versions = ">=3.8"
 files = [
-    {file = "jupyterlab-4.1.2-py3-none-any.whl", hash = "sha256:aa88193f03cf4d3555f6712f04d74112b5eb85edd7d222c588c7603a26d33c5b"},
-    {file = "jupyterlab-4.1.2.tar.gz", hash = "sha256:5d6348b3ed4085181499f621b7dfb6eb0b1f57f3586857aadfc8e3bf4c4885f9"},
+    {file = "jupyterlab-4.2.0-py3-none-any.whl", hash = "sha256:0dfe9278e25a145362289c555d9beb505697d269c10e99909766af7c440ad3cc"},
+    {file = "jupyterlab-4.2.0.tar.gz", hash = "sha256:356e9205a6a2ab689c47c8fe4919dba6c076e376d03f26baadc05748c2435dd5"},
 ]
 
 [package.dependencies]
@@ -1580,23 +1580,24 @@ async-lru = ">=1.0.0"
 httpx = ">=0.25.0"
 importlib-metadata = {version = ">=4.8.3", markers = "python_version < \"3.10\""}
 importlib-resources = {version = ">=1.4", markers = "python_version < \"3.9\""}
-ipykernel = "*"
+ipykernel = ">=6.5.0"
 jinja2 = ">=3.0.3"
 jupyter-core = "*"
 jupyter-lsp = ">=2.0.0"
 jupyter-server = ">=2.4.0,<3"
-jupyterlab-server = ">=2.19.0,<3"
+jupyterlab-server = ">=2.27.1,<3"
 notebook-shim = ">=0.2"
 packaging = "*"
-tomli = {version = "*", markers = "python_version < \"3.11\""}
+tomli = {version = ">=1.2.2", markers = "python_version < \"3.11\""}
 tornado = ">=6.2.0"
 traitlets = "*"
 
 [package.extras]
-dev = ["build", "bump2version", "coverage", "hatch", "pre-commit", "pytest-cov", "ruff (==0.2.0)"]
+dev = ["build", "bump2version", "coverage", "hatch", "pre-commit", "pytest-cov", "ruff (==0.3.5)"]
 docs = ["jsx-lexer", "myst-parser", "pydata-sphinx-theme (>=0.13.0)", "pytest", "pytest-check-links", "pytest-jupyter", "sphinx (>=1.8,<7.3.0)", "sphinx-copybutton"]
-docs-screenshots = ["altair (==5.2.0)", "ipython (==8.16.1)", "ipywidgets (==8.1.1)", "jupyterlab-geojson (==3.4.0)", "jupyterlab-language-pack-zh-cn (==4.0.post6)", "matplotlib (==3.8.2)", "nbconvert (>=7.0.0)", "pandas (==2.2.0)", "scipy (==1.12.0)", "vega-datasets (==0.9.0)"]
+docs-screenshots = ["altair (==5.3.0)", "ipython (==8.16.1)", "ipywidgets (==8.1.2)", "jupyterlab-geojson (==3.4.0)", "jupyterlab-language-pack-zh-cn (==4.1.post2)", "matplotlib (==3.8.3)", "nbconvert (>=7.0.0)", "pandas (==2.2.1)", "scipy (==1.12.0)", "vega-datasets (==0.9.0)"]
 test = ["coverage", "pytest (>=7.0)", "pytest-check-links (>=0.7)", "pytest-console-scripts", "pytest-cov", "pytest-jupyter (>=0.5.3)", "pytest-timeout", "pytest-tornasync", "requests", "requests-cache", "virtualenv"]
+upgrade-extension = ["copier (>=8,<10)", "jinja2-time (<0.3)", "pydantic (<2.0)", "pyyaml-include (<2.0)", "tomli-w (<2.0)"]
 
 [[package]]
 name = "jupyterlab-pygments"
@@ -1611,13 +1612,13 @@ files = [
 
 [[package]]
 name = "jupyterlab-server"
-version = "2.25.3"
+version = "2.27.1"
 description = "A set of server components for JupyterLab and JupyterLab like applications."
 optional = false
 python-versions = ">=3.8"
 files = [
-    {file = "jupyterlab_server-2.25.3-py3-none-any.whl", hash = "sha256:c48862519fded9b418c71645d85a49b2f0ec50d032ba8316738e9276046088c1"},
-    {file = "jupyterlab_server-2.25.3.tar.gz", hash = "sha256:846f125a8a19656611df5b03e5912c8393cea6900859baa64fa515eb64a8dc40"},
+    {file = "jupyterlab_server-2.27.1-py3-none-any.whl", hash = "sha256:f5e26156e5258b24d532c84e7c74cc212e203bff93eb856f81c24c16daeecc75"},
+    {file = "jupyterlab_server-2.27.1.tar.gz", hash = "sha256:097b5ac709b676c7284ac9c5e373f11930a561f52cd5a86e4fc7e5a9c8a8631d"},
 ]
 
 [package.dependencies]
@@ -1633,17 +1634,17 @@ requests = ">=2.31"
 [package.extras]
 docs = ["autodoc-traits", "jinja2 (<3.2.0)", "mistune (<4)", "myst-parser", "pydata-sphinx-theme", "sphinx", "sphinx-copybutton", "sphinxcontrib-openapi (>0.8)"]
 openapi = ["openapi-core (>=0.18.0,<0.19.0)", "ruamel-yaml"]
-test = ["hatch", "ipykernel", "openapi-core (>=0.18.0,<0.19.0)", "openapi-spec-validator (>=0.6.0,<0.8.0)", "pytest (>=7.0)", "pytest-console-scripts", "pytest-cov", "pytest-jupyter[server] (>=0.6.2)", "pytest-timeout", "requests-mock", "ruamel-yaml", "sphinxcontrib-spelling", "strict-rfc3339", "werkzeug"]
+test = ["hatch", "ipykernel", "openapi-core (>=0.18.0,<0.19.0)", "openapi-spec-validator (>=0.6.0,<0.8.0)", "pytest (>=7.0,<8)", "pytest-console-scripts", "pytest-cov", "pytest-jupyter[server] (>=0.6.2)", "pytest-timeout", "requests-mock", "ruamel-yaml", "sphinxcontrib-spelling", "strict-rfc3339", "werkzeug"]
 
 [[package]]
 name = "jupyterlab-widgets"
-version = "3.0.10"
+version = "3.0.11"
 description = "Jupyter interactive widgets for JupyterLab"
 optional = false
 python-versions = ">=3.7"
 files = [
-    {file = "jupyterlab_widgets-3.0.10-py3-none-any.whl", hash = "sha256:dd61f3ae7a5a7f80299e14585ce6cf3d6925a96c9103c978eda293197730cb64"},
-    {file = "jupyterlab_widgets-3.0.10.tar.gz", hash = "sha256:04f2ac04976727e4f9d0fa91cdc2f1ab860f965e504c29dbd6a65c882c9d04c0"},
+    {file = "jupyterlab_widgets-3.0.11-py3-none-any.whl", hash = "sha256:78287fd86d20744ace330a61625024cf5521e1c012a352ddc0a3cdc2348becd0"},
+    {file = "jupyterlab_widgets-3.0.11.tar.gz", hash = "sha256:dd5ac679593c969af29c9bed054c24f26842baa51352114736756bc035deee27"},
 ]
 
 [[package]]
@@ -2163,13 +2164,13 @@ requests = "*"
 
 [[package]]
 name = "mkdocs-git-revision-date-localized-plugin"
-version = "1.2.5"
+version = "1.2.6"
 description = "Mkdocs plugin that enables displaying the localized date of the last git modification of a markdown file."
 optional = false
 python-versions = ">=3.8"
 files = [
-    {file = "mkdocs_git_revision_date_localized_plugin-1.2.5-py3-none-any.whl", hash = "sha256:d796a18b07cfcdb154c133e3ec099d2bb5f38389e4fd54d3eb516a8a736815b8"},
-    {file = "mkdocs_git_revision_date_localized_plugin-1.2.5.tar.gz", hash = "sha256:0c439816d9d0dba48e027d9d074b2b9f1d7cd179f74ba46b51e4da7bb3dc4b9b"},
+    {file = "mkdocs_git_revision_date_localized_plugin-1.2.6-py3-none-any.whl", hash = "sha256:f015cb0f3894a39b33447b18e270ae391c4e25275cac5a626e80b243784e2692"},
+    {file = "mkdocs_git_revision_date_localized_plugin-1.2.6.tar.gz", hash = "sha256:e432942ce4ee8aa9b9f4493e993dee9d2cc08b3ea2b40a3d6b03ca0f2a4bcaa2"},
 ]
 
 [package.dependencies]
@@ -2198,13 +2199,13 @@ pygments = ">2.12.0"
 
 [[package]]
 name = "mkdocs-material"
-version = "9.5.22"
+version = "9.5.25"
 description = "Documentation that simply works"
 optional = false
 python-versions = ">=3.8"
 files = [
-    {file = "mkdocs_material-9.5.22-py3-none-any.whl", hash = "sha256:8c7a377d323567934e6cd46915e64dc209efceaec0dec1cf2202184f5649862c"},
-    {file = "mkdocs_material-9.5.22.tar.gz", hash = "sha256:22a853a456ae8c581c4628159574d6fc7c71b2c7569dc9c3a82cc70432219599"},
+    {file = "mkdocs_material-9.5.25-py3-none-any.whl", hash = "sha256:68fdab047a0b9bfbefe79ce267e8a7daaf5128bcf7867065fcd201ee335fece1"},
+    {file = "mkdocs_material-9.5.25.tar.gz", hash = "sha256:d0662561efb725b712207e0ee01f035ca15633f29a64628e24f01ec99d7078f4"},
 ]
 
 [package.dependencies]
@@ -2484,26 +2485,26 @@ setuptools = "*"
 
 [[package]]
 name = "notebook"
-version = "7.1.3"
+version = "7.2.0"
 description = "Jupyter Notebook - A web-based notebook environment for interactive computing"
 optional = false
 python-versions = ">=3.8"
 files = [
-    {file = "notebook-7.1.3-py3-none-any.whl", hash = "sha256:919b911e59f41f6e3857ce93c9d93535ba66bb090059712770e5968c07e1004d"},
-    {file = "notebook-7.1.3.tar.gz", hash = "sha256:41fcebff44cf7bb9377180808bcbae066629b55d8c7722f1ebbe75ca44f9cfc1"},
+    {file = "notebook-7.2.0-py3-none-any.whl", hash = "sha256:b4752d7407d6c8872fc505df0f00d3cae46e8efb033b822adacbaa3f1f3ce8f5"},
+    {file = "notebook-7.2.0.tar.gz", hash = "sha256:34a2ba4b08ad5d19ec930db7484fb79746a1784be9e1a5f8218f9af8656a141f"},
 ]
 
 [package.dependencies]
 jupyter-server = ">=2.4.0,<3"
-jupyterlab = ">=4.1.1,<4.2"
-jupyterlab-server = ">=2.22.1,<3"
+jupyterlab = ">=4.2.0,<4.3"
+jupyterlab-server = ">=2.27.1,<3"
 notebook-shim = ">=0.2,<0.3"
 tornado = ">=6.2.0"
 
 [package.extras]
 dev = ["hatch", "pre-commit"]
 docs = ["myst-parser", "nbsphinx", "pydata-sphinx-theme", "sphinx (>=1.3.6)", "sphinxcontrib-github-alt", "sphinxcontrib-spelling"]
-test = ["importlib-resources (>=5.0)", "ipykernel", "jupyter-server[test] (>=2.4.0,<3)", "jupyterlab-server[test] (>=2.22.1,<3)", "nbval", "pytest (>=7.0)", "pytest-console-scripts", "pytest-timeout", "pytest-tornasync", "requests"]
+test = ["importlib-resources (>=5.0)", "ipykernel", "jupyter-server[test] (>=2.4.0,<3)", "jupyterlab-server[test] (>=2.27.1,<3)", "nbval", "pytest (>=7.0)", "pytest-console-scripts", "pytest-timeout", "pytest-tornasync", "requests"]
 
 [[package]]
 name = "notebook-shim"
@@ -3038,13 +3039,13 @@ tomli = {version = ">=1.1.0", markers = "python_version < \"3.11\""}
 
 [[package]]
 name = "pytest"
-version = "8.2.0"
+version = "8.2.2"
 description = "pytest: simple powerful testing with Python"
 optional = false
 python-versions = ">=3.8"
 files = [
-    {file = "pytest-8.2.0-py3-none-any.whl", hash = "sha256:1733f0620f6cda4095bbf0d9ff8022486e91892245bb9e7d5542c018f612f233"},
-    {file = "pytest-8.2.0.tar.gz", hash = "sha256:d507d4482197eac0ba2bae2e9babf0672eb333017bcedaa5fb1a3d42c1174b3f"},
+    {file = "pytest-8.2.2-py3-none-any.whl", hash = "sha256:c434598117762e2bd304e526244f67bf66bbd7b5d6cf22138be51ff661980343"},
+    {file = "pytest-8.2.2.tar.gz", hash = "sha256:de4bb8104e201939ccdc688b27a89a7be2079b22e2bd2b07f806b6ba71117977"},
 ]
 
 [package.dependencies]
@@ -3460,13 +3461,13 @@ files = [
 
 [[package]]
 name = "requests"
-version = "2.31.0"
+version = "2.32.3"
 description = "Python HTTP for Humans."
 optional = false
-python-versions = ">=3.7"
+python-versions = ">=3.8"
 files = [
-    {file = "requests-2.31.0-py3-none-any.whl", hash = "sha256:58cd2187c01e70e6e26505bca751777aa9f2ee0b7f4300988b709f44e013003f"},
-    {file = "requests-2.31.0.tar.gz", hash = "sha256:942c5a758f98d790eaed1a29cb6eefc7ffb0d1cf7af05c3d2791656dbd6ad1e1"},
+    {file = "requests-2.32.3-py3-none-any.whl", hash = "sha256:70761cfe03c773ceb22aa2f671b4757976145175cdfca038c02654d061d6dcc6"},
+    {file = "requests-2.32.3.tar.gz", hash = "sha256:55365417734eb18255590a9ff9eb97e9e1da868d4ccd6402399eaf68af20a760"},
 ]
 
 [package.dependencies]
@@ -3661,28 +3662,28 @@ files = [
 
 [[package]]
 name = "ruff"
-version = "0.4.4"
+version = "0.4.7"
 description = "An extremely fast Python linter and code formatter, written in Rust."
 optional = false
 python-versions = ">=3.7"
 files = [
-    {file = "ruff-0.4.4-py3-none-macosx_10_12_x86_64.whl", hash = "sha256:29d44ef5bb6a08e235c8249294fa8d431adc1426bfda99ed493119e6f9ea1bf6"},
-    {file = "ruff-0.4.4-py3-none-macosx_11_0_arm64.whl", hash = "sha256:c4efe62b5bbb24178c950732ddd40712b878a9b96b1d02b0ff0b08a090cbd891"},
-    {file = "ruff-0.4.4-py3-none-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:4c8e2f1e8fc12d07ab521a9005d68a969e167b589cbcaee354cb61e9d9de9c15"},
-    {file = "ruff-0.4.4-py3-none-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:60ed88b636a463214905c002fa3eaab19795679ed55529f91e488db3fe8976ab"},
-    {file = "ruff-0.4.4-py3-none-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:b90fc5e170fc71c712cc4d9ab0e24ea505c6a9e4ebf346787a67e691dfb72e85"},
-    {file = "ruff-0.4.4-py3-none-manylinux_2_17_ppc64.manylinux2014_ppc64.whl", hash = "sha256:8e7e6ebc10ef16dcdc77fd5557ee60647512b400e4a60bdc4849468f076f6eef"},
-    {file = "ruff-0.4.4-py3-none-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:b9ddb2c494fb79fc208cd15ffe08f32b7682519e067413dbaf5f4b01a6087bcd"},
-    {file = "ruff-0.4.4-py3-none-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:c51c928a14f9f0a871082603e25a1588059b7e08a920f2f9fa7157b5bf08cfe9"},
-    {file = "ruff-0.4.4-py3-none-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:b5eb0a4bfd6400b7d07c09a7725e1a98c3b838be557fee229ac0f84d9aa49c36"},
-    {file = "ruff-0.4.4-py3-none-musllinux_1_2_aarch64.whl", hash = "sha256:b1867ee9bf3acc21778dcb293db504692eda5f7a11a6e6cc40890182a9f9e595"},
-    {file = "ruff-0.4.4-py3-none-musllinux_1_2_armv7l.whl", hash = "sha256:1aecced1269481ef2894cc495647392a34b0bf3e28ff53ed95a385b13aa45768"},
-    {file = "ruff-0.4.4-py3-none-musllinux_1_2_i686.whl", hash = "sha256:9da73eb616b3241a307b837f32756dc20a0b07e2bcb694fec73699c93d04a69e"},
-    {file = "ruff-0.4.4-py3-none-musllinux_1_2_x86_64.whl", hash = "sha256:958b4ea5589706a81065e2a776237de2ecc3e763342e5cc8e02a4a4d8a5e6f95"},
-    {file = "ruff-0.4.4-py3-none-win32.whl", hash = "sha256:cb53473849f011bca6e754f2cdf47cafc9c4f4ff4570003a0dad0b9b6890e876"},
-    {file = "ruff-0.4.4-py3-none-win_amd64.whl", hash = "sha256:424e5b72597482543b684c11def82669cc6b395aa8cc69acc1858b5ef3e5daae"},
-    {file = "ruff-0.4.4-py3-none-win_arm64.whl", hash = "sha256:39df0537b47d3b597293edbb95baf54ff5b49589eb7ff41926d8243caa995ea6"},
-    {file = "ruff-0.4.4.tar.gz", hash = "sha256:f87ea42d5cdebdc6a69761a9d0bc83ae9b3b30d0ad78952005ba6568d6c022af"},
+    {file = "ruff-0.4.7-py3-none-macosx_10_12_x86_64.whl", hash = "sha256:e089371c67892a73b6bb1525608e89a2aca1b77b5440acf7a71dda5dac958f9e"},
+    {file = "ruff-0.4.7-py3-none-macosx_11_0_arm64.whl", hash = "sha256:10f973d521d910e5f9c72ab27e409e839089f955be8a4c8826601a6323a89753"},
+    {file = "ruff-0.4.7-py3-none-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:59c3d110970001dfa494bcd95478e62286c751126dfb15c3c46e7915fc49694f"},
+    {file = "ruff-0.4.7-py3-none-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:fa9773c6c00f4958f73b317bc0fd125295110c3776089f6ef318f4b775f0abe4"},
+    {file = "ruff-0.4.7-py3-none-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:07fc80bbb61e42b3b23b10fda6a2a0f5a067f810180a3760c5ef1b456c21b9db"},
+    {file = "ruff-0.4.7-py3-none-manylinux_2_17_ppc64.manylinux2014_ppc64.whl", hash = "sha256:fa4dafe3fe66d90e2e2b63fa1591dd6e3f090ca2128daa0be33db894e6c18648"},
+    {file = "ruff-0.4.7-py3-none-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:a7c0083febdec17571455903b184a10026603a1de078428ba155e7ce9358c5f6"},
+    {file = "ruff-0.4.7-py3-none-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:ad1b20e66a44057c326168437d680a2166c177c939346b19c0d6b08a62a37589"},
+    {file = "ruff-0.4.7-py3-none-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:cbf5d818553add7511c38b05532d94a407f499d1a76ebb0cad0374e32bc67202"},
+    {file = "ruff-0.4.7-py3-none-musllinux_1_2_aarch64.whl", hash = "sha256:50e9651578b629baec3d1513b2534de0ac7ed7753e1382272b8d609997e27e83"},
+    {file = "ruff-0.4.7-py3-none-musllinux_1_2_armv7l.whl", hash = "sha256:8874a9df7766cb956b218a0a239e0a5d23d9e843e4da1e113ae1d27ee420877a"},
+    {file = "ruff-0.4.7-py3-none-musllinux_1_2_i686.whl", hash = "sha256:b9de9a6e49f7d529decd09381c0860c3f82fa0b0ea00ea78409b785d2308a567"},
+    {file = "ruff-0.4.7-py3-none-musllinux_1_2_x86_64.whl", hash = "sha256:13a1768b0691619822ae6d446132dbdfd568b700ecd3652b20d4e8bc1e498f78"},
+    {file = "ruff-0.4.7-py3-none-win32.whl", hash = "sha256:769e5a51df61e07e887b81e6f039e7ed3573316ab7dd9f635c5afaa310e4030e"},
+    {file = "ruff-0.4.7-py3-none-win_amd64.whl", hash = "sha256:9e3ab684ad403a9ed1226894c32c3ab9c2e0718440f6f50c7c5829932bc9e054"},
+    {file = "ruff-0.4.7-py3-none-win_arm64.whl", hash = "sha256:10f2204b9a613988e3484194c2c9e96a22079206b22b787605c255f130db5ed7"},
+    {file = "ruff-0.4.7.tar.gz", hash = "sha256:2331d2b051dc77a289a653fcc6a42cce357087c5975738157cd966590b18b5e1"},
 ]
 
 [[package]]
@@ -4009,13 +4010,13 @@ test = ["argcomplete (>=3.0.3)", "mypy (>=1.7.0)", "pre-commit", "pytest (>=7.0,
 
 [[package]]
 name = "twine"
-version = "5.0.0"
+version = "5.1.0"
 description = "Collection of utilities for publishing packages on PyPI"
 optional = false
 python-versions = ">=3.8"
 files = [
-    {file = "twine-5.0.0-py3-none-any.whl", hash = "sha256:a262933de0b484c53408f9edae2e7821c1c45a3314ff2df9bdd343aa7ab8edc0"},
-    {file = "twine-5.0.0.tar.gz", hash = "sha256:89b0cc7d370a4b66421cc6102f269aa910fe0f1861c124f573cf2ddedbc10cf4"},
+    {file = "twine-5.1.0-py3-none-any.whl", hash = "sha256:fe1d814395bfe50cfbe27783cb74efe93abeac3f66deaeb6c8390e4e92bacb43"},
+    {file = "twine-5.1.0.tar.gz", hash = "sha256:4d74770c88c4fcaf8134d2a6a9d863e40f08255ff7d8e2acb3cbbd57d25f6e9d"},
 ]
 
 [package.dependencies]
@@ -4226,13 +4227,13 @@ test = ["pytest (>=6.0.0)", "setuptools (>=65)"]
 
 [[package]]
 name = "widgetsnbextension"
-version = "4.0.10"
+version = "4.0.11"
 description = "Jupyter interactive widgets for Jupyter Notebook"
 optional = false
 python-versions = ">=3.7"
 files = [
-    {file = "widgetsnbextension-4.0.10-py3-none-any.whl", hash = "sha256:d37c3724ec32d8c48400a435ecfa7d3e259995201fbefa37163124a9fcb393cc"},
-    {file = "widgetsnbextension-4.0.10.tar.gz", hash = "sha256:64196c5ff3b9a9183a8e699a4227fb0b7002f252c814098e66c4d1cd0644688f"},
+    {file = "widgetsnbextension-4.0.11-py3-none-any.whl", hash = "sha256:55d4d6949d100e0d08b94948a42efc3ed6dfdc0e9468b2c4b128c9a2ce3a7a36"},
+    {file = "widgetsnbextension-4.0.11.tar.gz", hash = "sha256:8b22a8f1910bfd188e596fe7fc05dcbd87e810c8a4ba010bdb3da86637398474"},
 ]
 
 [[package]]
@@ -4257,4 +4258,4 @@ desktop = ["opencv-python"]
 [metadata]
 lock-version = "2.0"
 python-versions = "^3.8"
-content-hash = "29af5aa06f97e77a2dba94c5a6d77d7d1903448724df07416026a378d3c6a64d"
+content-hash = "e3d79f6c93041323b04c7b45e93bb3c4198b21889044004af8a0485a6145a207"

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "supervision"
-version = "0.21.0rc5"
+version = "0.21.0"
 description = "A set of easy-to-use utils that will come in handy in any Computer Vision project"
 authors = ["Piotr Skalski <piotr.skalski92@gmail.com>"]
 maintainers = ["Piotr Skalski <piotr.skalski92@gmail.com>"]
@@ -42,7 +42,7 @@ pyyaml = ">=5.3"
 defusedxml = "^0.7.1"
 opencv-python = { version = ">=4.5.5.64", optional = true }
 opencv-python-headless = ">=4.5.5.64"
-requests = { version = ">=2.26.0,<=2.31.0", optional = true }
+requests = { version = ">=2.26.0,<=2.32.3", optional = true }
 tqdm = { version = ">=4.62.3,<=4.66.4", optional = true }
 pillow = ">=9.4"
 

supervision/__init__.py

@@ -23,6 +23,7 @@ from supervision.annotators.core import (
     PercentageBarAnnotator,
     PixelateAnnotator,
     PolygonAnnotator,
+    RichLabelAnnotator,
     RoundBoxAnnotator,
     TraceAnnotator,
     TriangleAnnotator,
@@ -34,9 +35,17 @@ from supervision.dataset.core import (
     ClassificationDataset,
     DetectionDataset,
 )
+from supervision.dataset.utils import mask_to_rle, rle_to_mask
 from supervision.detection.annotate import BoxAnnotator
 from supervision.detection.core import Detections
 from supervision.detection.line_zone import LineZone, LineZoneAnnotator
+from supervision.detection.lmm import LMM
+from supervision.detection.overlap_filter import (
+    OverlapFilter,
+    box_non_max_merge,
+    box_non_max_suppression,
+    mask_non_max_suppression,
+)
 from supervision.detection.tools.csv_sink import CSVSink
 from supervision.detection.tools.inference_slicer import InferenceSlicer
 from supervision.detection.tools.json_sink import JSONSink
@@ -44,12 +53,12 @@ from supervision.detection.tools.polygon_zone import PolygonZone, PolygonZoneAnn
 from supervision.detection.tools.smoother import DetectionsSmoother
 from supervision.detection.utils import (
     box_iou_batch,
-    box_non_max_suppression,
     calculate_masks_centroids,
     clip_boxes,
+    contains_holes,
+    contains_multiple_segments,
     filter_polygons_by_area,
     mask_iou_batch,
-    mask_non_max_suppression,
     mask_to_polygons,
     mask_to_xyxy,
     move_boxes,

supervision/annotators/core.py

@@ -3,9 +3,15 @@ from typing import List, Optional, Tuple, Union
 
 import cv2
 import numpy as np
+from PIL import Image, ImageDraw, ImageFont
 
 from supervision.annotators.base import BaseAnnotator, ImageType
-from supervision.annotators.utils import ColorLookup, Trace, resolve_color
+from supervision.annotators.utils import (
+    ColorLookup,
+    Trace,
+    resolve_color,
+    resolve_text_background_xyxy,
+)
 from supervision.config import CLASS_NAME_DATA_FIELD, ORIENTED_BOX_COORDINATES
 from supervision.detection.core import Detections
 from supervision.detection.utils import clip_boxes, mask_to_polygons
@@ -936,59 +942,6 @@ class LabelAnnotator:
         self.text_anchor: Position = text_position
         self.color_lookup: ColorLookup = color_lookup
 
-    @staticmethod
-    def resolve_text_background_xyxy(
-        center_coordinates: Tuple[int, int],
-        text_wh: Tuple[int, int],
-        position: Position,
-    ) -> Tuple[int, int, int, int]:
-        center_x, center_y = center_coordinates
-        text_w, text_h = text_wh
-
-        if position == Position.TOP_LEFT:
-            return center_x, center_y - text_h, center_x + text_w, center_y
-        elif position == Position.TOP_RIGHT:
-            return center_x - text_w, center_y - text_h, center_x, center_y
-        elif position == Position.TOP_CENTER:
-            return (
-                center_x - text_w // 2,
-                center_y - text_h,
-                center_x + text_w // 2,
-                center_y,
-            )
-        elif position == Position.CENTER or position == Position.CENTER_OF_MASS:
-            return (
-                center_x - text_w // 2,
-                center_y - text_h // 2,
-                center_x + text_w // 2,
-                center_y + text_h // 2,
-            )
-        elif position == Position.BOTTOM_LEFT:
-            return center_x, center_y, center_x + text_w, center_y + text_h
-        elif position == Position.BOTTOM_RIGHT:
-            return center_x - text_w, center_y, center_x, center_y + text_h
-        elif position == Position.BOTTOM_CENTER:
-            return (
-                center_x - text_w // 2,
-                center_y,
-                center_x + text_w // 2,
-                center_y + text_h,
-            )
-        elif position == Position.CENTER_LEFT:
-            return (
-                center_x - text_w,
-                center_y - text_h // 2,
-                center_x,
-                center_y + text_h // 2,
-            )
-        elif position == Position.CENTER_RIGHT:
-            return (
-                center_x,
-                center_y - text_h // 2,
-                center_x + text_w,
-                center_y + text_h // 2,
-            )
-
     @convert_for_annotation_method
     def annotate(
         self,
@@ -1056,9 +1009,11 @@ class LabelAnnotator:
                 color=self.color,
                 detections=detections,
                 detection_idx=detection_idx,
-                color_lookup=self.color_lookup
-                if custom_color_lookup is None
-                else custom_color_lookup,
+                color_lookup=(
+                    self.color_lookup
+                    if custom_color_lookup is None
+                    else custom_color_lookup
+                ),
             )
 
             if labels is not None:
@@ -1078,7 +1033,7 @@ class LabelAnnotator:
             )[0]
             text_w_padded = text_w + 2 * self.text_padding
             text_h_padded = text_h + 2 * self.text_padding
-            text_background_xyxy = self.resolve_text_background_xyxy(
+            text_background_xyxy = resolve_text_background_xyxy(
                 center_coordinates=tuple(center_coordinates),
                 text_wh=(text_w_padded, text_h_padded),
                 position=self.text_anchor,
@@ -1148,6 +1103,165 @@ class LabelAnnotator:
         return scene
 
 
+class RichLabelAnnotator:
+    """
+    A class for annotating labels on an image using provided detections,
+    with support for Unicode characters by using a custom font.
+    """
+
+    def __init__(
+        self,
+        color: Union[Color, ColorPalette] = ColorPalette.DEFAULT,
+        text_color: Color = Color.WHITE,
+        font_path: str = None,
+        font_size: int = 10,
+        text_padding: int = 10,
+        text_position: Position = Position.TOP_LEFT,
+        color_lookup: ColorLookup = ColorLookup.CLASS,
+        border_radius: int = 0,
+    ):
+        """
+        Args:
+            color (Union[Color, ColorPalette]): The color or color palette to use for
+                annotating the text background.
+            text_color (Color): The color to use for the text.
+            font_path (str): Path to the font file (e.g., ".ttf" or ".otf") to use for
+                rendering text. If `None`, the default PIL font will be used.
+            font_size (int): Font size for the text.
+            text_padding (int): Padding around the text within its background box.
+            text_position (Position): Position of the text relative to the detection.
+                Possible values are defined in the `Position` enum.
+            color_lookup (ColorLookup): Strategy for mapping colors to annotations.
+                Options are `INDEX`, `CLASS`, `TRACK`.
+            border_radius (int): The radius to apply round edges. If the selected
+                value is higher than the lower dimension, width or height, is clipped.
+        """
+        self.color = color
+        self.text_color = text_color
+        self.text_padding = text_padding
+        self.text_anchor = text_position
+        self.color_lookup = color_lookup
+        self.border_radius = border_radius
+        if font_path is not None:
+            try:
+                self.font = ImageFont.truetype(font_path, font_size)
+            except OSError:
+                print(f"Font path '{font_path}' not found. Using PIL's default font.")
+                self.font = ImageFont.load_default(size=font_size)
+        else:
+            self.font = ImageFont.load_default(size=font_size)
+
+    def annotate(
+        self,
+        scene: ImageType,
+        detections: Detections,
+        labels: List[str] = None,
+        custom_color_lookup: Optional[np.ndarray] = None,
+    ) -> ImageType:
+        """
+        Annotates the given scene with labels based on the provided
+        detections, with support for Unicode characters.
+
+        Args:
+            scene (ImageType): The image where labels will be drawn.
+                `ImageType` is a flexible type, accepting either `numpy.ndarray`
+                or `PIL.Image.Image`.
+            detections (Detections): Object detections to annotate.
+            labels (List[str]): Optional. Custom labels for each detection.
+            custom_color_lookup (Optional[np.ndarray]): Custom color lookup array.
+                Allows to override the default color mapping strategy.
+
+        Returns:
+            The annotated image, matching the type of `scene` (`numpy.ndarray`
+                or `PIL.Image.Image`)
+
+        Example:
+            ```python
+            import supervision as sv
+
+            image = ...
+            detections = sv.Detections(...)
+
+            labels = [
+                f"{class_name} {confidence:.2f}"
+                for class_name, confidence
+                in zip(detections['class_name'], detections.confidence)
+            ]
+
+            rich_label_annotator = sv.RichLabelAnnotator(font_path="path/to/font.ttf")
+            annotated_frame = label_annotator.annotate(
+                scene=image.copy(),
+                detections=detections,
+                labels=labels
+            )
+            ```
+
+        """
+        if isinstance(scene, np.ndarray):
+            scene = Image.fromarray(cv2.cvtColor(scene, cv2.COLOR_BGR2RGB))
+        draw = ImageDraw.Draw(scene)
+        anchors_coordinates = detections.get_anchors_coordinates(
+            anchor=self.text_anchor
+        ).astype(int)
+        if labels is not None and len(labels) != len(detections):
+            raise ValueError(
+                f"The number of labels provided ({len(labels)}) does not match the "
+                f"number of detections ({len(detections)}). Each detection should have "
+                f"a corresponding label. This discrepancy can occur if the labels and "
+                f"detections are not aligned or if an incorrect number of labels has "
+                f"been provided. Please ensure that the labels array has the same "
+                f"length as the Detections object."
+            )
+        for detection_idx, center_coordinates in enumerate(anchors_coordinates):
+            color = resolve_color(
+                color=self.color,
+                detections=detections,
+                detection_idx=detection_idx,
+                color_lookup=(
+                    self.color_lookup
+                    if custom_color_lookup is None
+                    else custom_color_lookup
+                ),
+            )
+            if labels is not None:
+                text = labels[detection_idx]
+            elif detections[CLASS_NAME_DATA_FIELD] is not None:
+                text = detections[CLASS_NAME_DATA_FIELD][detection_idx]
+            elif detections.class_id is not None:
+                text = str(detections.class_id[detection_idx])
+            else:
+                text = str(detection_idx)
+
+            left, top, right, bottom = draw.textbbox((0, 0), text, font=self.font)
+            text_width = right - left
+            text_height = bottom - top
+            text_w_padded = text_width + 2 * self.text_padding
+            text_h_padded = text_height + 2 * self.text_padding
+            text_background_xyxy = resolve_text_background_xyxy(
+                center_coordinates=tuple(center_coordinates),
+                text_wh=(text_w_padded, text_h_padded),
+                position=self.text_anchor,
+            )
+
+            text_x = text_background_xyxy[0] + self.text_padding - left
+            text_y = text_background_xyxy[1] + self.text_padding - top
+
+            draw.rounded_rectangle(
+                text_background_xyxy,
+                radius=self.border_radius,
+                fill=color.as_rgb(),
+                outline=None,
+            )
+            draw.text(
+                xy=(text_x, text_y),
+                text=text,
+                font=self.font,
+                fill=self.text_color.as_rgb(),
+            )
+
+        return scene
+
+
 class BlurAnnotator(BaseAnnotator):
     """
     A class for blurring regions in an image using provided detections.

supervision/annotators/utils.py

@@ -1,5 +1,5 @@
 from enum import Enum
-from typing import Optional, Union
+from typing import Optional, Tuple, Union
 
 import numpy as np
 
@@ -34,14 +34,14 @@ def resolve_color_idx(
 ) -> int:
     if detection_idx >= len(detections):
         raise ValueError(
-            f"Detection index {detection_idx}"
+            f"Detection index {detection_idx} "
             f"is out of bounds for detections of length {len(detections)}"
         )
 
     if isinstance(color_lookup, np.ndarray):
         if len(color_lookup) != len(detections):
             raise ValueError(
-                f"Length of color lookup {len(color_lookup)}"
+                f"Length of color lookup {len(color_lookup)} "
                 f"does not match length of detections {len(detections)}"
             )
         return color_lookup[detection_idx]
@@ -50,19 +50,72 @@ def resolve_color_idx(
     elif color_lookup == ColorLookup.CLASS:
         if detections.class_id is None:
             raise ValueError(
-                "Could not resolve color by class because"
+                "Could not resolve color by class because "
                 "Detections do not have class_id"
             )
         return detections.class_id[detection_idx]
     elif color_lookup == ColorLookup.TRACK:
         if detections.tracker_id is None:
             raise ValueError(
-                "Could not resolve color by track because"
+                "Could not resolve color by track because "
                 "Detections do not have tracker_id"
             )
         return detections.tracker_id[detection_idx]
 
 
+def resolve_text_background_xyxy(
+    center_coordinates: Tuple[int, int],
+    text_wh: Tuple[int, int],
+    position: Position,
+) -> Tuple[int, int, int, int]:
+    center_x, center_y = center_coordinates
+    text_w, text_h = text_wh
+
+    if position == Position.TOP_LEFT:
+        return center_x, center_y - text_h, center_x + text_w, center_y
+    elif position == Position.TOP_RIGHT:
+        return center_x - text_w, center_y - text_h, center_x, center_y
+    elif position == Position.TOP_CENTER:
+        return (
+            center_x - text_w // 2,
+            center_y - text_h,
+            center_x + text_w // 2,
+            center_y,
+        )
+    elif position == Position.CENTER or position == Position.CENTER_OF_MASS:
+        return (
+            center_x - text_w // 2,
+            center_y - text_h // 2,
+            center_x + text_w // 2,
+            center_y + text_h // 2,
+        )
+    elif position == Position.BOTTOM_LEFT:
+        return center_x, center_y, center_x + text_w, center_y + text_h
+    elif position == Position.BOTTOM_RIGHT:
+        return center_x - text_w, center_y, center_x, center_y + text_h
+    elif position == Position.BOTTOM_CENTER:
+        return (
+            center_x - text_w // 2,
+            center_y,
+            center_x + text_w // 2,
+            center_y + text_h,
+        )
+    elif position == Position.CENTER_LEFT:
+        return (
+            center_x - text_w,
+            center_y - text_h // 2,
+            center_x,
+            center_y + text_h // 2,
+        )
+    elif position == Position.CENTER_RIGHT:
+        return (
+            center_x,
+            center_y - text_h // 2,
+            center_x + text_w,
+            center_y + text_h // 2,
+        )
+
+
 def get_color_by_index(color: Union[Color, ColorPalette], idx: int) -> Color:
     if isinstance(color, ColorPalette):
         return color.by_idx(idx)

supervision/dataset/core.py

@@ -116,13 +116,12 @@ class DetectionDataset(BaseDataset):
             Tuple[DetectionDataset, DetectionDataset]: A tuple containing
                 the training and testing datasets.
 
-        Example:
+        Examples:
             ```python
             import supervision as sv
 
             ds = sv.DetectionDataset(...)
-            train_ds, test_ds = ds.split(split_ratio=0.7,
-                                         random_state=42, shuffle=True)
+            train_ds, test_ds = ds.split(split_ratio=0.7, random_state=42, shuffle=True)
             len(train_ds), len(test_ds)
             # (700, 300)
             ```
@@ -229,7 +228,7 @@ class DetectionDataset(BaseDataset):
             DetectionDataset: A DetectionDataset instance containing
                 the loaded images and annotations.
 
-        Example:
+        Examples:
             ```python
             import roboflow
             from roboflow import Roboflow
@@ -286,7 +285,7 @@ class DetectionDataset(BaseDataset):
             DetectionDataset: A DetectionDataset instance
                 containing the loaded images and annotations.
 
-        Example:
+        Examples:
             ```python
             import roboflow
             from roboflow import Roboflow
@@ -391,7 +390,7 @@ class DetectionDataset(BaseDataset):
             DetectionDataset: A DetectionDataset instance containing
                 the loaded images and annotations.
 
-        Example:
+        Examples:
             ```python
             import roboflow
             from roboflow import Roboflow
@@ -431,6 +430,20 @@ class DetectionDataset(BaseDataset):
         Exports the dataset to COCO format. This method saves the
         images and their corresponding annotations in COCO format.
 
+        !!! tip
+
+            The format of the mask is determined automatically based on its structure:
+
+            - If a mask contains multiple disconnected components or holes, it will be
+            saved using the Run-Length Encoding (RLE) format for efficient storage and
+            processing.
+            - If a mask consists of a single, contiguous region without any holes, it
+            will be encoded as a polygon, preserving the outline of the object.
+
+            This automatic selection ensures that the masks are stored in the most
+            appropriate and space-efficient format, complying with COCO dataset
+            standards.
+
         Args:
             images_directory_path (Optional[str]): The path to the directory
                 where the images should be saved.
@@ -482,7 +495,7 @@ class DetectionDataset(BaseDataset):
             (DetectionDataset): A single `DetectionDataset` object containing
             the merged data from the input list.
 
-        Example:
+        Examples:
             ```python
             import supervision as sv
 
@@ -567,13 +580,12 @@ class ClassificationDataset(BaseDataset):
             Tuple[ClassificationDataset, ClassificationDataset]: A tuple containing
             the training and testing datasets.
 
-        Example:
+        Examples:
             ```python
             import supervision as sv
 
             cd = sv.ClassificationDataset(...)
-            train_cd,test_cd = cd.split(split_ratio=0.7,
-                                        random_state=42,shuffle=True)
+            train_cd,test_cd = cd.split(split_ratio=0.7, random_state=42,shuffle=True)
             len(train_cd), len(test_cd)
             # (700, 300)
             ```
@@ -635,7 +647,7 @@ class ClassificationDataset(BaseDataset):
         Returns:
             ClassificationDataset: The dataset.
 
-        Example:
+        Examples:
             ```python
             import roboflow
             from roboflow import Roboflow

supervision/dataset/formats/coco.py

@@ -5,13 +5,20 @@ from typing import Dict, List, Tuple
 
 import cv2
 import numpy as np
+import numpy.typing as npt
 
 from supervision.dataset.utils import (
     approximate_mask_with_polygons,
     map_detections_class_id,
+    mask_to_rle,
+    rle_to_mask,
 )
 from supervision.detection.core import Detections
-from supervision.detection.utils import polygon_to_mask
+from supervision.detection.utils import (
+    contains_holes,
+    contains_multiple_segments,
+    polygon_to_mask,
+)
 from supervision.utils.file import read_json_file, save_json_file
 
 
@@ -57,13 +64,24 @@ def group_coco_annotations_by_image_id(
     return annotations
 
 
-def _polygons_to_masks(
-    polygons: List[np.ndarray], resolution_wh: Tuple[int, int]
-) -> np.ndarray:
+def coco_annotations_to_masks(
+    image_annotations: List[dict], resolution_wh: Tuple[int, int]
+) -> npt.NDArray[np.bool_]:
     return np.array(
         [
-            polygon_to_mask(polygon=polygon, resolution_wh=resolution_wh)
-            for polygon in polygons
+            rle_to_mask(
+                rle=np.array(image_annotation["segmentation"]["counts"]),
+                resolution_wh=resolution_wh,
+            )
+            if image_annotation["iscrowd"]
+            else polygon_to_mask(
+                polygon=np.reshape(
+                    np.asarray(image_annotation["segmentation"], dtype=np.int32),
+                    (-1, 2),
+                ),
+                resolution_wh=resolution_wh,
+            )
+            for image_annotation in image_annotations
         ],
         dtype=bool,
     )
@@ -83,13 +101,9 @@ def coco_annotations_to_detections(
     xyxy[:, 2:4] += xyxy[:, 0:2]
 
     if with_masks:
-        polygons = [
-            np.reshape(
-                np.asarray(image_annotation["segmentation"], dtype=np.int32), (-1, 2)
-            )
-            for image_annotation in image_annotations
-        ]
-        mask = _polygons_to_masks(polygons=polygons, resolution_wh=resolution_wh)
+        mask = coco_annotations_to_masks(
+            image_annotations=image_annotations, resolution_wh=resolution_wh
+        )
         return Detections(
             class_id=np.asarray(class_ids, dtype=int), xyxy=xyxy, mask=mask
         )
@@ -108,24 +122,35 @@ def detections_to_coco_annotations(
     coco_annotations = []
     for xyxy, mask, _, class_id, _, _ in detections:
         box_width, box_height = xyxy[2] - xyxy[0], xyxy[3] - xyxy[1]
-        polygon = []
+        segmentation = []
+        iscrowd = 0
         if mask is not None:
-            polygon = list(
-                approximate_mask_with_polygons(
-                    mask=mask,
-                    min_image_area_percentage=min_image_area_percentage,
-                    max_image_area_percentage=max_image_area_percentage,
-                    approximation_percentage=approximation_percentage,
-                )[0].flatten()
-            )
+            iscrowd = contains_holes(mask=mask) or contains_multiple_segments(mask=mask)
+
+            if iscrowd:
+                segmentation = {
+                    "counts": mask_to_rle(mask=mask),
+                    "size": list(mask.shape[:2]),
+                }
+            else:
+                segmentation = [
+                    list(
+                        approximate_mask_with_polygons(
+                            mask=mask,
+                            min_image_area_percentage=min_image_area_percentage,
+                            max_image_area_percentage=max_image_area_percentage,
+                            approximation_percentage=approximation_percentage,
+                        )[0].flatten()
+                    )
+                ]
         coco_annotation = {
             "id": annotation_id,
             "image_id": image_id,
             "category_id": int(class_id),
             "bbox": [xyxy[0], xyxy[1], box_width, box_height],
             "area": box_width * box_height,
-            "segmentation": [polygon] if polygon else [],
-            "iscrowd": 0,
+            "segmentation": segmentation,
+            "iscrowd": iscrowd,
         }
         coco_annotations.append(coco_annotation)
         annotation_id += 1

supervision/dataset/utils.py

@@ -2,10 +2,11 @@ import copy
 import os
 import random
 from pathlib import Path
-from typing import Dict, List, Optional, Tuple, TypeVar
+from typing import Dict, List, Optional, Tuple, TypeVar, Union
 
 import cv2
 import numpy as np
+import numpy.typing as npt
 
 from supervision.detection.core import Detections
 from supervision.detection.utils import (
@@ -129,3 +130,123 @@ def train_test_split(
 
     split_index = int(len(data) * train_ratio)
     return data[:split_index], data[split_index:]
+
+
+def rle_to_mask(
+    rle: Union[npt.NDArray[np.int_], List[int]], resolution_wh: Tuple[int, int]
+) -> npt.NDArray[np.bool_]:
+    """
+    Converts run-length encoding (RLE) to a binary mask.
+
+    Args:
+        rle (Union[npt.NDArray[np.int_], List[int]]): The 1D RLE array, the format
+            used in the COCO dataset (column-wise encoding, values of an array with
+            even indices represent the number of pixels assigned as background,
+            values of an array with odd indices represent the number of pixels
+            assigned as foreground object).
+        resolution_wh (Tuple[int, int]): The width (w) and height (h)
+            of the desired binary mask.
+
+    Returns:
+        The generated 2D Boolean mask of shape `(h, w)`, where the foreground object is
+            marked with `True`'s and the rest is filled with `False`'s.
+
+    Raises:
+        AssertionError: If the sum of pixels encoded in RLE differs from the
+            number of pixels in the expected mask (computed based on resolution_wh).
+
+    Examples:
+        ```python
+        import supervision as sv
+
+        sv.rle_to_mask([5, 2, 2, 2, 5], (4, 4))
+        # array([
+        #     [False, False, False, False],
+        #     [False, True,  True,  False],
+        #     [False, True,  True,  False],
+        #     [False, False, False, False],
+        # ])
+        ```
+    """
+    if isinstance(rle, list):
+        rle = np.array(rle, dtype=int)
+
+    width, height = resolution_wh
+
+    assert width * height == np.sum(rle), (
+        "the sum of the number of pixels in the RLE must be the same "
+        "as the number of pixels in the expected mask"
+    )
+
+    zero_one_values = np.zeros(shape=(rle.size, 1), dtype=np.uint8)
+    zero_one_values[1::2] = 1
+
+    decoded_rle = np.repeat(zero_one_values, rle, axis=0)
+    decoded_rle = np.append(
+        decoded_rle, np.zeros(width * height - len(decoded_rle), dtype=np.uint8)
+    )
+    return decoded_rle.reshape((height, width), order="F")
+
+
+def mask_to_rle(mask: npt.NDArray[np.bool_]) -> List[int]:
+    """
+    Converts a binary mask into a run-length encoding (RLE).
+
+    Args:
+        mask (npt.NDArray[np.bool_]): 2D binary mask where `True` indicates foreground
+            object and `False` indicates background.
+
+    Returns:
+        The run-length encoded mask. Values of a list with even indices
+            represent the number of pixels assigned as background (`False`), values
+            of a list with odd indices represent the number of pixels assigned
+            as foreground object (`True`).
+
+    Raises:
+        AssertionError: If input mask is not 2D or is empty.
+
+    Examples:
+        ```python
+        import numpy as np
+        import supervision as sv
+
+        mask = np.array([
+            [True, True, True, True],
+            [True, True, True, True],
+            [True, True, True, True],
+            [True, True, True, True],
+        ])
+        sv.mask_to_rle(mask)
+        # [0, 16]
+
+        mask = np.array([
+            [False, False, False, False],
+            [False, True,  True,  False],
+            [False, True,  True,  False],
+            [False, False, False, False],
+        ])
+        sv.mask_to_rle(mask)
+        # [5, 2, 2, 2, 5]
+        ```
+
+    ![mask_to_rle](https://media.roboflow.com/supervision-docs/mask-to-rle.png){ align=center width="800" }
+    """  # noqa E501 // docs
+    assert mask.ndim == 2, "Input mask must be 2D"
+    assert mask.size != 0, "Input mask cannot be empty"
+
+    on_value_change_indices = np.where(
+        mask.ravel(order="F") != np.roll(mask.ravel(order="F"), 1)
+    )[0]
+
+    on_value_change_indices = np.append(on_value_change_indices, mask.size)
+    # need to add 0 at the beginning when the same value is in the first and
+    # last element of the flattened mask
+    if on_value_change_indices[0] != 0:
+        on_value_change_indices = np.insert(on_value_change_indices, 0, 0)
+
+    rle = np.diff(on_value_change_indices)
+
+    if mask[0][0] == 1:
+        rle = np.insert(rle, 0, 0)
+
+    return list(rle)

supervision/detection/core.py

@@ -7,20 +7,25 @@ from typing import Any, Dict, Iterator, List, Optional, Tuple, Union
 import numpy as np
 
 from supervision.config import CLASS_NAME_DATA_FIELD, ORIENTED_BOX_COORDINATES
-from supervision.detection.utils import (
+from supervision.detection.lmm import LMM, from_paligemma, validate_lmm_and_kwargs
+from supervision.detection.overlap_filter import (
+    box_non_max_merge,
     box_non_max_suppression,
+    mask_non_max_suppression,
+)
+from supervision.detection.utils import (
+    box_iou_batch,
     calculate_masks_centroids,
     extract_ultralytics_masks,
     get_data_item,
     is_data_equal,
-    mask_non_max_suppression,
     mask_to_xyxy,
     merge_data,
     process_roboflow_result,
     xywh_to_xyxy,
 )
 from supervision.geometry.core import Position
-from supervision.utils.internal import deprecated
+from supervision.utils.internal import deprecated, get_instance_variables
 from supervision.validators import validate_detections_fields
 
 
@@ -240,7 +245,7 @@ class Detections:
             Class names values can be accessed using `detections["class_name"]`.
         """  # noqa: E501 // docs
 
-        if "obb" in ultralytics_results and ultralytics_results.obb is not None:
+        if hasattr(ultralytics_results, "obb") and ultralytics_results.obb is not None:
             class_id = ultralytics_results.obb.cls.cpu().numpy().astype(int)
             class_names = np.array([ultralytics_results.names[i] for i in class_id])
             oriented_box_coordinates = ultralytics_results.obb.xyxyxyxy.cpu().numpy()
@@ -805,6 +810,52 @@ class Detections:
             class_id=paddledet_result["bbox"][:, 0].astype(int),
         )
 
+    @classmethod
+    def from_lmm(cls, lmm: Union[LMM, str], result: str, **kwargs) -> Detections:
+        """
+        Creates a Detections object from the given result string based on the specified
+        Large Multimodal Model (LMM).
+
+        Args:
+            lmm (Union[LMM, str]): The type of LMM (Large Multimodal Model) to use.
+            result (str): The result string containing the detection data.
+            **kwargs: Additional keyword arguments required by the specified LMM.
+
+        Returns:
+            Detections: A new Detections object.
+
+        Raises:
+            ValueError: If the LMM is invalid, required arguments are missing, or
+                disallowed arguments are provided.
+            ValueError: If the specified LMM is not supported.
+
+        Examples:
+            ```python
+            import supervision as sv
+
+            paligemma_result = "<loc0256><loc0256><loc0768><loc0768> cat"
+            detections = sv.Detections.from_lmm(
+                sv.LMM.PALIGEMMA,
+                paligemma_result,
+                resolution_wh=(1000, 1000),
+                classes=['cat', 'dog']
+            )
+            detections.xyxy
+            # array([[250., 250., 750., 750.]])
+
+            detections.class_id
+            # array([0])
+            ```
+        """
+        lmm = validate_lmm_and_kwargs(lmm, kwargs)
+
+        if lmm == LMM.PALIGEMMA:
+            xyxy, class_id, class_name = from_paligemma(result, **kwargs)
+            data = {CLASS_NAME_DATA_FIELD: class_name}
+            return cls(xyxy=xyxy, class_id=class_id, data=data)
+
+        raise ValueError(f"Unsupported LMM: {lmm}")
+
     @classmethod
     def empty(cls) -> Detections:
         """
@@ -827,6 +878,14 @@ class Detections:
             class_id=np.array([], dtype=int),
         )
 
+    def is_empty(self) -> bool:
+        """
+        Returns `True` if the `Detections` object is considered empty.
+        """
+        empty_detections = Detections.empty()
+        empty_detections.data = self.data
+        return self == empty_detections
+
     @classmethod
     def merge(cls, detections_list: List[Detections]) -> Detections:
         """
@@ -839,6 +898,10 @@ class Detections:
         For example, if merging Detections with 3 and 4 detected objects, this method
         will return a Detections with 7 objects (7 entries in `xyxy`, `mask`, etc).
 
+        !!! Note
+
+            When merging, empty `Detections` objects are ignored.
+
         Args:
             detections_list (List[Detections]): A list of Detections objects to merge.
 
@@ -877,6 +940,10 @@ class Detections:
             array([0.1, 0.2, 0.3])
             ```
         """
+        detections_list = [
+            detections for detections in detections_list if not detections.is_empty()
+        ]
+
         if len(detections_list) == 0:
             return Detections.empty()
 
@@ -895,12 +962,13 @@ class Detections:
         def stack_or_none(name: str):
             if all(d.__getattribute__(name) is None for d in detections_list):
                 return None
-            stack_list = [
-                d.__getattribute__(name)
-                for d in detections_list
-                if d.__getattribute__(name) is not None
-            ]
-            return np.vstack(stack_list) if name == "mask" else np.hstack(stack_list)
+            if any(d.__getattribute__(name) is None for d in detections_list):
+                raise ValueError(f"All or none of the '{name}' fields must be None")
+            return (
+                np.vstack([d.__getattribute__(name) for d in detections_list])
+                if name == "mask"
+                else np.hstack([d.__getattribute__(name) for d in detections_list])
+            )
 
         mask = stack_or_none("mask")
         confidence = stack_or_none("confidence")
@@ -1150,3 +1218,195 @@ class Detections:
             )
 
         return self[indices]
+
+    def with_nmm(
+        self, threshold: float = 0.5, class_agnostic: bool = False
+    ) -> Detections:
+        """
+        Perform non-maximum merging on the current set of object detections.
+
+        Args:
+            threshold (float, optional): The intersection-over-union threshold
+                to use for non-maximum merging. Defaults to 0.5.
+            class_agnostic (bool, optional): Whether to perform class-agnostic
+                non-maximum merging. If True, the class_id of each detection
+                will be ignored. Defaults to False.
+
+        Returns:
+            Detections: A new Detections object containing the subset of detections
+                after non-maximum merging.
+
+        Raises:
+            AssertionError: If `confidence` is None or `class_id` is None and
+                class_agnostic is False.
+
+        ![non-max-merging](https://media.roboflow.com/supervision-docs/non-max-merging.png){ align=center width="800" }
+        """  # noqa: E501 // docs
+        if len(self) == 0:
+            return self
+
+        assert (
+            self.confidence is not None
+        ), "Detections confidence must be given for NMM to be executed."
+
+        if class_agnostic:
+            predictions = np.hstack((self.xyxy, self.confidence.reshape(-1, 1)))
+        else:
+            assert self.class_id is not None, (
+                "Detections class_id must be given for NMM to be executed. If you"
+                " intended to perform class agnostic NMM set class_agnostic=True."
+            )
+            predictions = np.hstack(
+                (
+                    self.xyxy,
+                    self.confidence.reshape(-1, 1),
+                    self.class_id.reshape(-1, 1),
+                )
+            )
+
+        merge_groups = box_non_max_merge(
+            predictions=predictions, iou_threshold=threshold
+        )
+
+        result = []
+        for merge_group in merge_groups:
+            unmerged_detections = [self[i] for i in merge_group]
+            merged_detections = merge_inner_detections_objects(
+                unmerged_detections, threshold
+            )
+            result.append(merged_detections)
+
+        return Detections.merge(result)
+
+
+def merge_inner_detection_object_pair(
+    detections_1: Detections, detections_2: Detections
+) -> Detections:
+    """
+    Merges two Detections object into a single Detections object.
+    Assumes each Detections contains exactly one object.
+
+    A `winning` detection is determined based on the confidence score of the two
+    input detections. This winning detection is then used to specify which
+    `class_id`, `tracker_id`, and `data` to include in the merged Detections object.
+
+    The resulting `confidence` of the merged object is calculated by the weighted
+    contribution of ea detection to the merged object.
+    The bounding boxes and masks of the two input detections are merged into a
+    single bounding box and mask, respectively.
+
+    Args:
+        detections_1 (Detections):
+            The first Detections object
+        detections_2 (Detections):
+            The second Detections object
+
+    Returns:
+        Detections: A new Detections object, with merged attributes.
+
+    Raises:
+        ValueError: If the input Detections objects do not have exactly 1 detected
+            object.
+
+    Example:
+        ```python
+        import cv2
+        import supervision as sv
+        from inference import get_model
+
+        image = cv2.imread(<SOURCE_IMAGE_PATH>)
+        model = get_model(model_id="yolov8s-640")
+
+        result = model.infer(image)[0]
+        detections = sv.Detections.from_inference(result)
+
+        merged_detections = merge_object_detection_pair(
+            detections[0], detections[1])
+        ```
+    """
+    if len(detections_1) != 1 or len(detections_2) != 1:
+        raise ValueError("Both Detections should have exactly 1 detected object.")
+
+    validate_fields_both_defined_or_none(detections_1, detections_2)
+
+    xyxy_1 = detections_1.xyxy[0]
+    xyxy_2 = detections_2.xyxy[0]
+    if detections_1.confidence is None and detections_2.confidence is None:
+        merged_confidence = None
+    else:
+        detection_1_area = (xyxy_1[2] - xyxy_1[0]) * (xyxy_1[3] - xyxy_1[1])
+        detections_2_area = (xyxy_2[2] - xyxy_2[0]) * (xyxy_2[3] - xyxy_2[1])
+        merged_confidence = (
+            detection_1_area * detections_1.confidence[0]
+            + detections_2_area * detections_2.confidence[0]
+        ) / (detection_1_area + detections_2_area)
+        merged_confidence = np.array([merged_confidence])
+
+    merged_x1, merged_y1 = np.minimum(xyxy_1[:2], xyxy_2[:2])
+    merged_x2, merged_y2 = np.maximum(xyxy_1[2:], xyxy_2[2:])
+    merged_xyxy = np.array([[merged_x1, merged_y1, merged_x2, merged_y2]])
+
+    if detections_1.mask is None and detections_2.mask is None:
+        merged_mask = None
+    else:
+        merged_mask = np.logical_or(detections_1.mask, detections_2.mask)
+
+    if detections_1.confidence is None and detections_2.confidence is None:
+        winning_detection = detections_1
+    elif detections_1.confidence[0] >= detections_2.confidence[0]:
+        winning_detection = detections_1
+    else:
+        winning_detection = detections_2
+
+    return Detections(
+        xyxy=merged_xyxy,
+        mask=merged_mask,
+        confidence=merged_confidence,
+        class_id=winning_detection.class_id,
+        tracker_id=winning_detection.tracker_id,
+        data=winning_detection.data,
+    )
+
+
+def merge_inner_detections_objects(
+    detections: List[Detections], threshold=0.5
+) -> Detections:
+    """
+    Given N detections each of length 1 (exactly one object inside), combine them into a
+    single detection object of length 1. The contained inner object will be the merged
+    result of all the input detections.
+
+    For example, this lets you merge N boxes into one big box, N masks into one mask,
+    etc.
+    """
+    detections_1 = detections[0]
+    for detections_2 in detections[1:]:
+        box_iou = box_iou_batch(detections_1.xyxy, detections_2.xyxy)[0]
+        if box_iou < threshold:
+            break
+        detections_1 = merge_inner_detection_object_pair(detections_1, detections_2)
+    return detections_1
+
+
+def validate_fields_both_defined_or_none(
+    detections_1: Detections, detections_2: Detections
+) -> None:
+    """
+    Verify that for each optional field in the Detections, both instances either have
+    the field set to None or both have it set to non-None values.
+
+    `data` field is ignored.
+
+    Raises:
+        ValueError: If one field is None and the other is not, for any of the fields.
+    """
+    attributes = get_instance_variables(detections_1)
+    for attribute in attributes:
+        value_1 = getattr(detections_1, attribute)
+        value_2 = getattr(detections_2, attribute)
+
+        if (value_1 is None) != (value_2 is None):
+            raise ValueError(
+                f"Field '{attribute}' should be consistently None or not None in both "
+                "Detections."
+            )

supervision/detection/line_zone.py

@@ -1,12 +1,15 @@
+import warnings
 from typing import Dict, Iterable, Optional, Tuple
 
 import cv2
 import numpy as np
 
 from supervision.detection.core import Detections
+from supervision.detection.utils import cross_product
 from supervision.draw.color import Color
 from supervision.draw.utils import draw_text
 from supervision.geometry.core import Point, Position, Vector
+from supervision.utils.internal import SupervisionWarnings
 
 
 class LineZone:
@@ -81,6 +84,8 @@ class LineZone:
         self.in_count: int = 0
         self.out_count: int = 0
         self.triggering_anchors = triggering_anchors
+        if not list(self.triggering_anchors):
+            raise ValueError("Triggering anchors cannot be empty.")
 
     @staticmethod
     def calculate_region_of_interest_limits(vector: Vector) -> Tuple[Vector, Vector]:
@@ -140,6 +145,15 @@ class LineZone:
         if len(detections) == 0:
             return crossed_in, crossed_out
 
+        if detections.tracker_id is None:
+            warnings.warn(
+                "Line zone counting skipped. LineZone requires tracker_id. Refer to "
+                "https://supervision.roboflow.com/latest/trackers for more "
+                "information.",
+                category=SupervisionWarnings,
+            )
+            return crossed_in, crossed_out
+
         all_anchors = np.array(
             [
                 detections.get_anchors_coordinates(anchor)
@@ -147,31 +161,23 @@ class LineZone:
             ]
         )
 
+        cross_products_1 = cross_product(all_anchors, self.limits[0])
+        cross_products_2 = cross_product(all_anchors, self.limits[1])
+        in_limits = (cross_products_1 > 0) == (cross_products_2 > 0)
+        in_limits = np.all(in_limits, axis=0)
+
+        triggers = cross_product(all_anchors, self.vector) < 0
+        has_any_left_trigger = np.any(triggers, axis=0)
+        has_any_right_trigger = np.any(~triggers, axis=0)
+        is_uniformly_triggered = ~(has_any_left_trigger & has_any_right_trigger)
         for i, tracker_id in enumerate(detections.tracker_id):
-            if tracker_id is None:
+            if not in_limits[i]:
                 continue
 
-            box_anchors = [Point(x=x, y=y) for x, y in all_anchors[:, i, :]]
-
-            in_limits = all(
-                [
-                    self.is_point_in_limits(point=anchor, limits=self.limits)
-                    for anchor in box_anchors
-                ]
-            )
-
-            if not in_limits:
+            if not is_uniformly_triggered[i]:
                 continue
 
-            triggers = [
-                self.vector.cross_product(point=anchor) < 0 for anchor in box_anchors
-            ]
-
-            if len(set(triggers)) == 2:
-                continue
-
-            tracker_state = triggers[0]
-
+            tracker_state = has_any_left_trigger[i]
             if tracker_id not in self.tracker_state:
                 self.tracker_state[tracker_id] = tracker_state
                 continue

supervision/detection/lmm.py

@@ -0,0 +1,59 @@
+import re
+from enum import Enum
+from typing import Any, Dict, List, Optional, Tuple, Union
+
+import numpy as np
+
+
+class LMM(Enum):
+    PALIGEMMA = "paligemma"
+
+
+REQUIRED_ARGUMENTS: Dict[LMM, List[str]] = {LMM.PALIGEMMA: ["resolution_wh"]}
+
+ALLOWED_ARGUMENTS: Dict[LMM, List[str]] = {LMM.PALIGEMMA: ["resolution_wh", "classes"]}
+
+
+def validate_lmm_and_kwargs(lmm: Union[LMM, str], kwargs: Dict[str, Any]) -> LMM:
+    if isinstance(lmm, str):
+        try:
+            lmm = LMM(lmm.lower())
+        except ValueError:
+            raise ValueError(
+                f"Invalid lmm value: {lmm}. Must be one of {[e.value for e in LMM]}"
+            )
+
+    required_args = REQUIRED_ARGUMENTS.get(lmm, [])
+    for arg in required_args:
+        if arg not in kwargs:
+            raise ValueError(f"Missing required argument: {arg}")
+
+    allowed_args = ALLOWED_ARGUMENTS.get(lmm, [])
+    for arg in kwargs:
+        if arg not in allowed_args:
+            raise ValueError(f"Argument {arg} is not allowed for {lmm.name}")
+
+    return lmm
+
+
+def from_paligemma(
+    result: str, resolution_wh: Tuple[int, int], classes: Optional[List[str]] = None
+) -> Tuple[np.ndarray, Optional[np.ndarray], np.ndarray]:
+    w, h = resolution_wh
+    pattern = re.compile(
+        r"(?<!<loc\d{4}>)<loc(\d{4})><loc(\d{4})><loc(\d{4})><loc(\d{4})> ([\w\s\-]+)"
+    )
+    matches = pattern.findall(result)
+    matches = np.array(matches) if matches else np.empty((0, 5))
+
+    xyxy, class_name = matches[:, [1, 0, 3, 2]], matches[:, 4]
+    xyxy = xyxy.astype(int) / 1024 * np.array([w, h, w, h])
+    class_name = np.char.strip(class_name.astype(str))
+    class_id = None
+
+    if classes is not None:
+        mask = np.array([name in classes for name in class_name]).astype(bool)
+        xyxy, class_name = xyxy[mask], class_name[mask]
+        class_id = np.array([classes.index(name) for name in class_name])
+
+    return xyxy, class_id, class_name

supervision/detection/overlap_filter.py

@@ -0,0 +1,263 @@
+from enum import Enum
+from typing import List, Union
+
+import numpy as np
+import numpy.typing as npt
+
+from supervision.detection.utils import box_iou_batch, mask_iou_batch
+
+
+def resize_masks(masks: np.ndarray, max_dimension: int = 640) -> np.ndarray:
+    """
+    Resize all masks in the array to have a maximum dimension of max_dimension,
+    maintaining aspect ratio.
+
+    Args:
+        masks (np.ndarray): 3D array of binary masks with shape (N, H, W).
+        max_dimension (int): The maximum dimension for the resized masks.
+
+    Returns:
+        np.ndarray: Array of resized masks.
+    """
+    max_height = np.max(masks.shape[1])
+    max_width = np.max(masks.shape[2])
+    scale = min(max_dimension / max_height, max_dimension / max_width)
+
+    new_height = int(scale * max_height)
+    new_width = int(scale * max_width)
+
+    x = np.linspace(0, max_width - 1, new_width).astype(int)
+    y = np.linspace(0, max_height - 1, new_height).astype(int)
+    xv, yv = np.meshgrid(x, y)
+
+    resized_masks = masks[:, yv, xv]
+
+    resized_masks = resized_masks.reshape(masks.shape[0], new_height, new_width)
+    return resized_masks
+
+
+def mask_non_max_suppression(
+    predictions: np.ndarray,
+    masks: np.ndarray,
+    iou_threshold: float = 0.5,
+    mask_dimension: int = 640,
+) -> np.ndarray:
+    """
+    Perform Non-Maximum Suppression (NMS) on segmentation predictions.
+
+    Args:
+        predictions (np.ndarray): A 2D array of object detection predictions in
+            the format of `(x_min, y_min, x_max, y_max, score)`
+            or `(x_min, y_min, x_max, y_max, score, class)`. Shape: `(N, 5)` or
+            `(N, 6)`, where N is the number of predictions.
+        masks (np.ndarray): A 3D array of binary masks corresponding to the predictions.
+            Shape: `(N, H, W)`, where N is the number of predictions, and H, W are the
+            dimensions of each mask.
+        iou_threshold (float, optional): The intersection-over-union threshold
+            to use for non-maximum suppression.
+        mask_dimension (int, optional): The dimension to which the masks should be
+            resized before computing IOU values. Defaults to 640.
+
+    Returns:
+        np.ndarray: A boolean array indicating which predictions to keep after
+            non-maximum suppression.
+
+    Raises:
+        AssertionError: If `iou_threshold` is not within the closed
+        range from `0` to `1`.
+    """
+    assert 0 <= iou_threshold <= 1, (
+        "Value of `iou_threshold` must be in the closed range from 0 to 1, "
+        f"{iou_threshold} given."
+    )
+    rows, columns = predictions.shape
+
+    if columns == 5:
+        predictions = np.c_[predictions, np.zeros(rows)]
+
+    sort_index = predictions[:, 4].argsort()[::-1]
+    predictions = predictions[sort_index]
+    masks = masks[sort_index]
+    masks_resized = resize_masks(masks, mask_dimension)
+    ious = mask_iou_batch(masks_resized, masks_resized)
+    categories = predictions[:, 5]
+
+    keep = np.ones(rows, dtype=bool)
+    for i in range(rows):
+        if keep[i]:
+            condition = (ious[i] > iou_threshold) & (categories[i] == categories)
+            keep[i + 1 :] = np.where(condition[i + 1 :], False, keep[i + 1 :])
+
+    return keep[sort_index.argsort()]
+
+
+def box_non_max_suppression(
+    predictions: np.ndarray, iou_threshold: float = 0.5
+) -> np.ndarray:
+    """
+    Perform Non-Maximum Suppression (NMS) on object detection predictions.
+
+    Args:
+        predictions (np.ndarray): An array of object detection predictions in
+            the format of `(x_min, y_min, x_max, y_max, score)`
+            or `(x_min, y_min, x_max, y_max, score, class)`.
+        iou_threshold (float, optional): The intersection-over-union threshold
+            to use for non-maximum suppression.
+
+    Returns:
+        np.ndarray: A boolean array indicating which predictions to keep after n
+            on-maximum suppression.
+
+    Raises:
+        AssertionError: If `iou_threshold` is not within the
+            closed range from `0` to `1`.
+    """
+    assert 0 <= iou_threshold <= 1, (
+        "Value of `iou_threshold` must be in the closed range from 0 to 1, "
+        f"{iou_threshold} given."
+    )
+    rows, columns = predictions.shape
+
+    # add column #5 - category filled with zeros for agnostic nms
+    if columns == 5:
+        predictions = np.c_[predictions, np.zeros(rows)]
+
+    # sort predictions column #4 - score
+    sort_index = np.flip(predictions[:, 4].argsort())
+    predictions = predictions[sort_index]
+
+    boxes = predictions[:, :4]
+    categories = predictions[:, 5]
+    ious = box_iou_batch(boxes, boxes)
+    ious = ious - np.eye(rows)
+
+    keep = np.ones(rows, dtype=bool)
+
+    for index, (iou, category) in enumerate(zip(ious, categories)):
+        if not keep[index]:
+            continue
+
+        # drop detections with iou > iou_threshold and
+        # same category as current detections
+        condition = (iou > iou_threshold) & (categories == category)
+        keep = keep & ~condition
+
+    return keep[sort_index.argsort()]
+
+
+def group_overlapping_boxes(
+    predictions: npt.NDArray[np.float64], iou_threshold: float = 0.5
+) -> List[List[int]]:
+    """
+    Apply greedy version of non-maximum merging to avoid detecting too many
+    overlapping bounding boxes for a given object.
+
+    Args:
+        predictions (npt.NDArray[np.float64]): An array of shape `(n, 5)` containing
+            the bounding boxes coordinates in format `[x1, y1, x2, y2]`
+            and the confidence scores.
+        iou_threshold (float, optional): The intersection-over-union threshold
+            to use for non-maximum suppression. Defaults to 0.5.
+
+    Returns:
+        List[List[int]]: Groups of prediction indices be merged.
+            Each group may have 1 or more elements.
+    """
+    merge_groups: List[List[int]] = []
+
+    scores = predictions[:, 4]
+    order = scores.argsort()
+
+    while len(order) > 0:
+        idx = int(order[-1])
+
+        order = order[:-1]
+        if len(order) == 0:
+            merge_groups.append([idx])
+            break
+
+        merge_candidate = np.expand_dims(predictions[idx], axis=0)
+        ious = box_iou_batch(predictions[order][:, :4], merge_candidate[:, :4])
+        ious = ious.flatten()
+
+        above_threshold = ious >= iou_threshold
+        merge_group = [idx] + np.flip(order[above_threshold]).tolist()
+        merge_groups.append(merge_group)
+        order = order[~above_threshold]
+    return merge_groups
+
+
+def box_non_max_merge(
+    predictions: npt.NDArray[np.float64],
+    iou_threshold: float = 0.5,
+) -> List[List[int]]:
+    """
+    Apply greedy version of non-maximum merging per category to avoid detecting
+    too many overlapping bounding boxes for a given object.
+
+    Args:
+        predictions (npt.NDArray[np.float64]): An array of shape `(n, 5)` or `(n, 6)`
+            containing the bounding boxes coordinates in format `[x1, y1, x2, y2]`,
+            the confidence scores and class_ids. Omit class_id column to allow
+            detections of different classes to be merged.
+        iou_threshold (float, optional): The intersection-over-union threshold
+            to use for non-maximum suppression. Defaults to 0.5.
+
+    Returns:
+        List[List[int]]: Groups of prediction indices be merged.
+            Each group may have 1 or more elements.
+    """
+    if predictions.shape[1] == 5:
+        return group_overlapping_boxes(predictions, iou_threshold)
+
+    category_ids = predictions[:, 5]
+    merge_groups = []
+    for category_id in np.unique(category_ids):
+        curr_indices = np.where(category_ids == category_id)[0]
+        merge_class_groups = group_overlapping_boxes(
+            predictions[curr_indices], iou_threshold
+        )
+
+        for merge_class_group in merge_class_groups:
+            merge_groups.append(curr_indices[merge_class_group].tolist())
+
+    for merge_group in merge_groups:
+        if len(merge_group) == 0:
+            raise ValueError(
+                f"Empty group detected when non-max-merging "
+                f"detections: {merge_groups}"
+            )
+    return merge_groups
+
+
+class OverlapFilter(Enum):
+    """
+    Enum specifying the strategy for filtering overlapping detections.
+
+    Attributes:
+        NONE: Do not filter detections based on overlap.
+        NON_MAX_SUPPRESSION: Filter detections using non-max suppression. This means,
+            detections that overlap by more than a set threshold will be discarded,
+            except for the one with the highest confidence.
+        NON_MAX_MERGE: Merge detections with non-max merging. This means,
+            detections that overlap by more than a set threshold will be merged
+            into a single detection.
+    """
+
+    NONE = "none"
+    NON_MAX_SUPPRESSION = "non_max_suppression"
+    NON_MAX_MERGE = "non_max_merge"
+
+
+def validate_overlap_filter(
+    strategy: Union[OverlapFilter, str],
+) -> OverlapFilter:
+    if isinstance(strategy, str):
+        try:
+            strategy = OverlapFilter(strategy.lower())
+        except ValueError:
+            raise ValueError(
+                f"Invalid strategy value: {strategy}. Must be one of "
+                f"{[e.value for e in OverlapFilter]}"
+            )
+    return strategy

supervision/detection/tools/inference_slicer.py

@@ -1,11 +1,14 @@
+import warnings
 from concurrent.futures import ThreadPoolExecutor, as_completed
-from typing import Callable, Optional, Tuple
+from typing import Callable, Optional, Tuple, Union
 
 import numpy as np
 
 from supervision.detection.core import Detections
+from supervision.detection.overlap_filter import OverlapFilter, validate_overlap_filter
 from supervision.detection.utils import move_boxes, move_masks
 from supervision.utils.image import crop_image
+from supervision.utils.internal import SupervisionWarnings
 
 
 def move_detections(
@@ -50,8 +53,10 @@ class InferenceSlicer:
             `(width, height)`.
         overlap_ratio_wh (Tuple[float, float]): Overlap ratio between consecutive
             slices in the format `(width_ratio, height_ratio)`.
-        iou_threshold (Optional[float]): Intersection over Union (IoU) threshold
-            used for non-max suppression.
+        overlap_filter_strategy (Union[OverlapFilter, str]): Strategy for
+            filtering or merging overlapping detections in slices.
+        iou_threshold (float): Intersection over Union (IoU) threshold
+            used when filtering by overlap.
         callback (Callable): A function that performs inference on a given image
             slice and returns detections.
         thread_workers (int): Number of threads for parallel execution.
@@ -68,12 +73,18 @@ class InferenceSlicer:
         callback: Callable[[np.ndarray], Detections],
         slice_wh: Tuple[int, int] = (320, 320),
         overlap_ratio_wh: Tuple[float, float] = (0.2, 0.2),
-        iou_threshold: Optional[float] = 0.5,
+        overlap_filter_strategy: Union[
+            OverlapFilter, str
+        ] = OverlapFilter.NON_MAX_SUPPRESSION,
+        iou_threshold: float = 0.5,
         thread_workers: int = 1,
     ):
+        overlap_filter_strategy = validate_overlap_filter(overlap_filter_strategy)
+
         self.slice_wh = slice_wh
         self.overlap_ratio_wh = overlap_ratio_wh
         self.iou_threshold = iou_threshold
+        self.overlap_filter_strategy = overlap_filter_strategy
         self.callback = callback
         self.thread_workers = thread_workers
 
@@ -104,7 +115,10 @@ class InferenceSlicer:
                 result = model(image_slice)[0]
                 return sv.Detections.from_ultralytics(result)
 
-            slicer = sv.InferenceSlicer(callback = callback)
+            slicer = sv.InferenceSlicer(
+                callback=callback,
+                overlap_filter_strategy=sv.OverlapFilter.NON_MAX_SUPPRESSION,
+            )
 
             detections = slicer(image)
             ```
@@ -124,9 +138,19 @@ class InferenceSlicer:
             for future in as_completed(futures):
                 detections_list.append(future.result())
 
-        return Detections.merge(detections_list=detections_list).with_nms(
-            threshold=self.iou_threshold
-        )
+        merged = Detections.merge(detections_list=detections_list)
+        if self.overlap_filter_strategy == OverlapFilter.NONE:
+            return merged
+        elif self.overlap_filter_strategy == OverlapFilter.NON_MAX_SUPPRESSION:
+            return merged.with_nms(threshold=self.iou_threshold)
+        elif self.overlap_filter_strategy == OverlapFilter.NON_MAX_MERGE:
+            return merged.with_nmm(threshold=self.iou_threshold)
+        else:
+            warnings.warn(
+                f"Invalid overlap filter strategy: {self.overlap_filter_strategy}",
+                category=SupervisionWarnings,
+            )
+            return merged
 
     def _run_callback(self, image, offset) -> Detections:
         """

supervision/detection/tools/polygon_zone.py

@@ -54,6 +54,8 @@ class PolygonZone:
 
         self.polygon = polygon.astype(int)
         self.triggering_anchors = triggering_anchors
+        if not list(self.triggering_anchors):
+            raise ValueError("Triggering anchors cannot be empty.")
 
         self.current_count = 0
 

supervision/detection/tools/smoother.py

@@ -1,3 +1,4 @@
+import warnings
 from collections import defaultdict, deque
 from copy import deepcopy
 from typing import Optional
@@ -5,6 +6,7 @@ from typing import Optional
 import numpy as np
 
 from supervision.detection.core import Detections
+from supervision.utils.internal import SupervisionWarnings
 
 
 class DetectionsSmoother:
@@ -70,16 +72,16 @@ class DetectionsSmoother:
         """
 
         if detections.tracker_id is None:
-            print(
+            warnings.warn(
                 "Smoothing skipped. DetectionsSmoother requires tracker_id. Refer to "
-                "https://supervision.roboflow.com/latest/trackers for more information."
+                "https://supervision.roboflow.com/latest/trackers for more "
+                "information.",
+                category=SupervisionWarnings,
             )
             return detections
 
         for detection_idx in range(len(detections)):
             tracker_id = detections.tracker_id[detection_idx]
-            if tracker_id is None:
-                continue
 
             self.tracks[tracker_id].append(detections[detection_idx])
 

supervision/detection/utils.py

@@ -3,8 +3,10 @@ from typing import Dict, List, Optional, Tuple, Union
 
 import cv2
 import numpy as np
+import numpy.typing as npt
 
 from supervision.config import CLASS_NAME_DATA_FIELD
+from supervision.geometry.core import Vector
 
 MIN_POLYGON_POINT_COUNT = 3
 
@@ -56,7 +58,9 @@ def box_iou_batch(boxes_true: np.ndarray, boxes_detection: np.ndarray) -> np.nda
     bottom_right = np.minimum(boxes_true[:, None, 2:], boxes_detection[:, 2:])
 
     area_inter = np.prod(np.clip(bottom_right - top_left, a_min=0, a_max=None), 2)
-    return area_inter / (area_true[:, None] + area_detection - area_inter)
+    ious = area_inter / (area_true[:, None] + area_detection - area_inter)
+    ious = np.nan_to_num(ious)
+    return ious
 
 
 def _mask_iou_batch_split(
@@ -136,144 +140,6 @@ def mask_iou_batch(
     return np.vstack(ious)
 
 
-def resize_masks(masks: np.ndarray, max_dimension: int = 640) -> np.ndarray:
-    """
-    Resize all masks in the array to have a maximum dimension of max_dimension,
-    maintaining aspect ratio.
-
-    Args:
-        masks (np.ndarray): 3D array of binary masks with shape (N, H, W).
-        max_dimension (int): The maximum dimension for the resized masks.
-
-    Returns:
-        np.ndarray: Array of resized masks.
-    """
-    max_height = np.max(masks.shape[1])
-    max_width = np.max(masks.shape[2])
-    scale = min(max_dimension / max_height, max_dimension / max_width)
-
-    new_height = int(scale * max_height)
-    new_width = int(scale * max_width)
-
-    x = np.linspace(0, max_width - 1, new_width).astype(int)
-    y = np.linspace(0, max_height - 1, new_height).astype(int)
-    xv, yv = np.meshgrid(x, y)
-
-    resized_masks = masks[:, yv, xv]
-
-    resized_masks = resized_masks.reshape(masks.shape[0], new_height, new_width)
-    return resized_masks
-
-
-def mask_non_max_suppression(
-    predictions: np.ndarray,
-    masks: np.ndarray,
-    iou_threshold: float = 0.5,
-    mask_dimension: int = 640,
-) -> np.ndarray:
-    """
-    Perform Non-Maximum Suppression (NMS) on segmentation predictions.
-
-    Args:
-        predictions (np.ndarray): A 2D array of object detection predictions in
-            the format of `(x_min, y_min, x_max, y_max, score)`
-            or `(x_min, y_min, x_max, y_max, score, class)`. Shape: `(N, 5)` or
-            `(N, 6)`, where N is the number of predictions.
-        masks (np.ndarray): A 3D array of binary masks corresponding to the predictions.
-            Shape: `(N, H, W)`, where N is the number of predictions, and H, W are the
-            dimensions of each mask.
-        iou_threshold (float, optional): The intersection-over-union threshold
-            to use for non-maximum suppression.
-        mask_dimension (int, optional): The dimension to which the masks should be
-            resized before computing IOU values. Defaults to 640.
-
-    Returns:
-        np.ndarray: A boolean array indicating which predictions to keep after
-            non-maximum suppression.
-
-    Raises:
-        AssertionError: If `iou_threshold` is not within the closed
-        range from `0` to `1`.
-    """
-    assert 0 <= iou_threshold <= 1, (
-        "Value of `iou_threshold` must be in the closed range from 0 to 1, "
-        f"{iou_threshold} given."
-    )
-    rows, columns = predictions.shape
-
-    if columns == 5:
-        predictions = np.c_[predictions, np.zeros(rows)]
-
-    sort_index = predictions[:, 4].argsort()[::-1]
-    predictions = predictions[sort_index]
-    masks = masks[sort_index]
-    masks_resized = resize_masks(masks, mask_dimension)
-    ious = mask_iou_batch(masks_resized, masks_resized)
-    categories = predictions[:, 5]
-
-    keep = np.ones(rows, dtype=bool)
-    for i in range(rows):
-        if keep[i]:
-            condition = (ious[i] > iou_threshold) & (categories[i] == categories)
-            keep[i + 1 :] = np.where(condition[i + 1 :], False, keep[i + 1 :])
-
-    return keep[sort_index.argsort()]
-
-
-def box_non_max_suppression(
-    predictions: np.ndarray, iou_threshold: float = 0.5
-) -> np.ndarray:
-    """
-    Perform Non-Maximum Suppression (NMS) on object detection predictions.
-
-    Args:
-        predictions (np.ndarray): An array of object detection predictions in
-            the format of `(x_min, y_min, x_max, y_max, score)`
-            or `(x_min, y_min, x_max, y_max, score, class)`.
-        iou_threshold (float, optional): The intersection-over-union threshold
-            to use for non-maximum suppression.
-
-    Returns:
-        np.ndarray: A boolean array indicating which predictions to keep after n
-            on-maximum suppression.
-
-    Raises:
-        AssertionError: If `iou_threshold` is not within the
-            closed range from `0` to `1`.
-    """
-    assert 0 <= iou_threshold <= 1, (
-        "Value of `iou_threshold` must be in the closed range from 0 to 1, "
-        f"{iou_threshold} given."
-    )
-    rows, columns = predictions.shape
-
-    # add column #5 - category filled with zeros for agnostic nms
-    if columns == 5:
-        predictions = np.c_[predictions, np.zeros(rows)]
-
-    # sort predictions column #4 - score
-    sort_index = np.flip(predictions[:, 4].argsort())
-    predictions = predictions[sort_index]
-
-    boxes = predictions[:, :4]
-    categories = predictions[:, 5]
-    ious = box_iou_batch(boxes, boxes)
-    ious = ious - np.eye(rows)
-
-    keep = np.ones(rows, dtype=bool)
-
-    for index, (iou, category) in enumerate(zip(ious, categories)):
-        if not keep[index]:
-            continue
-
-        # drop detections with iou > iou_threshold and
-        # same category as current detections
-        condition = (iou > iou_threshold) & (categories == category)
-        keep = keep & ~condition
-
-    return keep[sort_index.argsort()]
-
-
 def clip_boxes(xyxy: np.ndarray, resolution_wh: Tuple[int, int]) -> np.ndarray:
     """
     Clips bounding boxes coordinates to fit within the frame resolution.
@@ -289,6 +155,25 @@ def clip_boxes(xyxy: np.ndarray, resolution_wh: Tuple[int, int]) -> np.ndarray:
         np.ndarray: A numpy array of shape `(N, 4)` where each row
             corresponds to a bounding box with coordinates clipped to fit
             within the frame resolution.
+
+    Examples:
+        ```python
+        import numpy as np
+        import supervision as sv
+
+        xyxy = np.array([
+            [10, 20, 300, 200],
+            [15, 25, 350, 450],
+            [-10, -20, 30, 40]
+        ])
+
+        sv.clip_boxes(xyxy=xyxy, resolution_wh=(320, 240))
+        # array([
+        #     [ 10,  20, 300, 200],
+        #     [ 15,  25, 320, 240],
+        #     [  0,   0,  30,  40]
+        # ])
+        ```
     """
     result = np.copy(xyxy)
     width, height = resolution_wh
@@ -315,6 +200,23 @@ def pad_boxes(xyxy: np.ndarray, px: int, py: Optional[int] = None) -> np.ndarray
         np.ndarray: A numpy array of shape `(N, 4)` where each row corresponds to a
             bounding box with coordinates padded according to the provided padding
             values.
+
+    Examples:
+        ```python
+        import numpy as np
+        import supervision as sv
+
+        xyxy = np.array([
+            [10, 20, 30, 40],
+            [15, 25, 35, 45]
+        ])
+
+        sv.pad_boxes(xyxy=xyxy, px=5, py=10)
+        # array([
+        #     [ 5, 10, 35, 50],
+        #     [10, 15, 40, 55]
+        # ])
+        ```
     """
     if py is None:
         py = px
@@ -346,7 +248,7 @@ def mask_to_xyxy(masks: np.ndarray) -> np.ndarray:
             `(x_min, y_min, x_max, y_max)` for each mask
     """
     n = masks.shape[0]
-    bboxes = np.zeros((n, 4), dtype=int)
+    xyxy = np.zeros((n, 4), dtype=int)
 
     for i, mask in enumerate(masks):
         rows, cols = np.where(mask)
@@ -354,9 +256,9 @@ def mask_to_xyxy(masks: np.ndarray) -> np.ndarray:
         if len(rows) > 0 and len(cols) > 0:
             x_min, x_max = np.min(cols), np.max(cols)
             y_min, y_max = np.min(rows), np.max(rows)
-            bboxes[i, :] = [x_min, y_min, x_max, y_max]
+            xyxy[i, :] = [x_min, y_min, x_max, y_max]
 
-    return bboxes
+    return xyxy
 
 
 def mask_to_polygons(mask: np.ndarray) -> List[np.ndarray]:
@@ -592,29 +494,34 @@ def process_roboflow_result(
     return xyxy, confidence, class_id, masks, tracker_id, data
 
 
-def move_boxes(xyxy: np.ndarray, offset: np.ndarray) -> np.ndarray:
+def move_boxes(
+    xyxy: npt.NDArray[np.float64], offset: npt.NDArray[np.int32]
+) -> npt.NDArray[np.float64]:
     """
     Parameters:
-        xyxy (np.ndarray): An array of shape `(n, 4)` containing the bounding boxes
-            coordinates in format `[x1, y1, x2, y2]`
+        xyxy (npt.NDArray[np.float64]): An array of shape `(n, 4)` containing the
+            bounding boxes coordinates in format `[x1, y1, x2, y2]`
         offset (np.array): An array of shape `(2,)` containing offset values in format
             is `[dx, dy]`.
 
     Returns:
-        np.ndarray: Repositioned bounding boxes.
+        npt.NDArray[np.float64]: Repositioned bounding boxes.
 
-    Example:
+    Examples:
         ```python
         import numpy as np
         import supervision as sv
 
-        boxes = np.array([[10, 10, 20, 20], [30, 30, 40, 40]])
+        xyxy = np.array([
+            [10, 10, 20, 20],
+            [30, 30, 40, 40]
+        ])
         offset = np.array([5, 5])
-        moved_box = sv.move_boxes(boxes, offset)
-        print(moved_box)
-        # np.array([
+
+        sv.move_boxes(xyxy=xyxy, offset=offset)
+        # array([
         #    [15, 15, 25, 25],
-        #     [35, 35, 45, 45]
+        #    [35, 35, 45, 45]
         # ])
         ```
     """
@@ -622,24 +529,25 @@ def move_boxes(xyxy: np.ndarray, offset: np.ndarray) -> np.ndarray:
 
 
 def move_masks(
-    masks: np.ndarray,
-    offset: np.ndarray,
-    resolution_wh: Tuple[int, int] = None,
-) -> np.ndarray:
+    masks: npt.NDArray[np.bool_],
+    offset: npt.NDArray[np.int32],
+    resolution_wh: Tuple[int, int],
+) -> npt.NDArray[np.bool_]:
     """
     Offset the masks in an array by the specified (x, y) amount.
 
     Args:
-        masks (np.ndarray): A 3D array of binary masks corresponding to the predictions.
-            Shape: `(N, H, W)`, where N is the number of predictions, and H, W are the
-            dimensions of each mask.
-        offset (np.ndarray): An array of shape `(2,)` containing non-negative int values
-            `[dx, dy]`.
+        masks (npt.NDArray[np.bool_]): A 3D array of binary masks corresponding to the
+            predictions. Shape: `(N, H, W)`, where N is the number of predictions, and
+            H, W are the dimensions of each mask.
+        offset (npt.NDArray[np.int32]): An array of shape `(2,)` containing non-negative
+            int values `[dx, dy]`.
         resolution_wh (Tuple[int, int]): The width and height of the desired mask
             resolution.
 
     Returns:
-        (np.ndarray) repositioned masks, optionally padded to the specified shape.
+        (npt.NDArray[np.bool_]) repositioned masks, optionally padded to the specified
+            shape.
     """
 
     if offset[0] < 0 or offset[1] < 0:
@@ -655,30 +563,34 @@ def move_masks(
     return mask_array
 
 
-def scale_boxes(xyxy: np.ndarray, factor: float) -> np.ndarray:
+def scale_boxes(
+    xyxy: npt.NDArray[np.float64], factor: float
+) -> npt.NDArray[np.float64]:
     """
     Scale the dimensions of bounding boxes.
 
     Parameters:
-        xyxy (np.ndarray): An array of shape `(n, 4)` containing the bounding boxes
-            coordinates in format `[x1, y1, x2, y2]`
+        xyxy (npt.NDArray[np.float64]): An array of shape `(n, 4)` containing the
+            bounding boxes coordinates in format `[x1, y1, x2, y2]`
         factor (float): A float value representing the factor by which the box
             dimensions are scaled. A factor greater than 1 enlarges the boxes, while a
             factor less than 1 shrinks them.
 
     Returns:
-        np.ndarray: Scaled bounding boxes.
+        npt.NDArray[np.float64]: Scaled bounding boxes.
 
-    Example:
+    Examples:
         ```python
         import numpy as np
         import supervision as sv
 
-        boxes = np.array([[10, 10, 20, 20], [30, 30, 40, 40]])
-        factor = 1.5
-        scaled_bb = sv.scale_boxes(boxes, factor)
-        print(scaled_bb)
-        # np.array([
+        xyxy = np.array([
+            [10, 10, 20, 20],
+            [30, 30, 40, 40]
+        ])
+
+        sv.scale_boxes(xyxy=xyxy, factor=1.5)
+        # array([
         #    [ 7.5,  7.5, 22.5, 22.5],
         #    [27.5, 27.5, 42.5, 42.5]
         # ])
@@ -735,19 +647,19 @@ def is_data_equal(data_a: Dict[str, np.ndarray], data_b: Dict[str, np.ndarray])
 
 
 def merge_data(
-    data_list: List[Dict[str, Union[np.ndarray, List]]],
-) -> Dict[str, Union[np.ndarray, List]]:
+    data_list: List[Dict[str, Union[npt.NDArray[np.generic], List]]],
+) -> Dict[str, Union[npt.NDArray[np.generic], List]]:
     """
     Merges the data payloads of a list of Detections instances.
 
     Args:
         data_list: The data payloads of the Detections instances. Each data payload
             is a dictionary with the same keys, and the values are either lists or
-            np.ndarray.
+            npt.NDArray[np.generic].
 
     Returns:
         A single data payload containing the merged data, preserving the original data
-            types (list or np.ndarray).
+            types (list or npt.NDArray[np.generic]).
 
     Raises:
         ValueError: If data values within a single object have different lengths or if
@@ -756,6 +668,10 @@ def merge_data(
     if not data_list:
         return {}
 
+    all_keys_sets = [set(data.keys()) for data in data_list]
+    if not all(keys_set == all_keys_sets[0] for keys_set in all_keys_sets):
+        raise ValueError("All data dictionaries must have the same keys to merge.")
+
     for data in data_list:
         lengths = [len(value) for value in data.values()]
         if len(set(lengths)) > 1:
@@ -763,21 +679,7 @@ def merge_data(
                 "All data values within a single object must have equal length."
             )
 
-    keys_by_data = [set(data.keys()) for data in data_list]
-    keys_by_data = [keys for keys in keys_by_data if len(keys) > 0]
-    if not keys_by_data:
-        return {}
-
-    common_keys = set.intersection(*keys_by_data)
-    all_keys = set.union(*keys_by_data)
-    if common_keys != all_keys:
-        raise ValueError(
-            f"All sv.Detections.data dictionaries must have the same keys. Common "
-            f"keys: {common_keys}, but some dictionaries have additional keys: "
-            f"{all_keys.difference(common_keys)}."
-        )
-
-    merged_data = {key: [] for key in all_keys}
+    merged_data = {key: [] for key in all_keys_sets[0]}
     for data in data_list:
         for key in data:
             merged_data[key].append(data[key])
@@ -840,3 +742,138 @@ def get_data_item(
             raise TypeError(f"Unsupported data type for key '{key}': {type(value)}")
 
     return subset_data
+
+
+def contains_holes(mask: npt.NDArray[np.bool_]) -> bool:
+    """
+    Checks if the binary mask contains holes (background pixels fully enclosed by
+    foreground pixels).
+
+    Args:
+        mask (npt.NDArray[np.bool_]): 2D binary mask where `True` indicates foreground
+            object and `False` indicates background.
+
+    Returns:
+        True if holes are detected, False otherwise.
+
+    Examples:
+        ```python
+        import numpy as np
+        import supervision as sv
+
+        mask = np.array([
+            [0, 0, 0, 0, 0],
+            [0, 1, 1, 1, 0],
+            [0, 1, 0, 1, 0],
+            [0, 1, 1, 1, 0],
+            [0, 0, 0, 0, 0]
+        ]).astype(bool)
+
+        sv.contains_holes(mask=mask)
+        # True
+
+        mask = np.array([
+            [0, 0, 0, 0, 0],
+            [0, 1, 1, 1, 0],
+            [0, 1, 1, 1, 0],
+            [0, 1, 1, 1, 0],
+            [0, 0, 0, 0, 0]
+        ]).astype(bool)
+
+        sv.contains_holes(mask=mask)
+        # False
+        ```
+
+    ![contains_holes](https://media.roboflow.com/supervision-docs/contains-holes.png){ align=center width="800" }
+    """  # noqa E501 // docs
+    mask_uint8 = mask.astype(np.uint8)
+    _, hierarchy = cv2.findContours(mask_uint8, cv2.RETR_CCOMP, cv2.CHAIN_APPROX_SIMPLE)
+
+    if hierarchy is not None:
+        parent_contour_index = 3
+        for h in hierarchy[0]:
+            if h[parent_contour_index] != -1:
+                return True
+    return False
+
+
+def contains_multiple_segments(
+    mask: npt.NDArray[np.bool_], connectivity: int = 4
+) -> bool:
+    """
+    Checks if the binary mask contains multiple unconnected foreground segments.
+
+    Args:
+        mask (npt.NDArray[np.bool_]): 2D binary mask where `True` indicates foreground
+            object and `False` indicates background.
+        connectivity (int) : Default: 4 is 4-way connectivity, which means that
+            foreground pixels are the part of the same segment/component
+            if their edges touch.
+            Alternatively: 8 for 8-way connectivity, when foreground pixels are
+            connected by their edges or corners touch.
+
+    Returns:
+        True when the mask contains multiple not connected components, False otherwise.
+
+    Raises:
+        ValueError: If connectivity(int) parameter value is not 4 or 8.
+
+    Examples:
+        ```python
+        import numpy as np
+        import supervision as sv
+
+        mask = np.array([
+            [0, 0, 0, 0, 0, 0],
+            [0, 1, 1, 0, 1, 1],
+            [0, 1, 1, 0, 1, 1],
+            [0, 0, 0, 0, 0, 0],
+            [0, 1, 1, 1, 0, 0],
+            [0, 1, 1, 1, 0, 0]
+        ]).astype(bool)
+
+        sv.contains_multiple_segments(mask=mask, connectivity=4)
+        # True
+
+        mask = np.array([
+            [0, 0, 0, 0, 0, 0],
+            [0, 1, 1, 1, 1, 1],
+            [0, 1, 1, 1, 1, 1],
+            [0, 1, 1, 1, 1, 1],
+            [0, 1, 1, 1, 1, 1],
+            [0, 0, 0, 0, 0, 0]
+        ]).astype(bool)
+
+        sv.contains_multiple_segments(mask=mask, connectivity=4)
+        # False
+        ```
+
+    ![contains_multiple_segments](https://media.roboflow.com/supervision-docs/contains-multiple-segments.png){ align=center width="800" }
+    """  # noqa E501 // docs
+    if connectivity != 4 and connectivity != 8:
+        raise ValueError(
+            "Incorrect connectivity value. Possible connectivity values: 4 or 8."
+        )
+    mask_uint8 = mask.astype(np.uint8)
+    labels = np.zeros_like(mask_uint8, dtype=np.int32)
+    number_of_labels, _ = cv2.connectedComponents(
+        mask_uint8, labels, connectivity=connectivity
+    )
+    return number_of_labels > 2
+
+
+def cross_product(anchors: np.ndarray, vector: Vector) -> np.ndarray:
+    """
+    Get array of cross products of each anchor with a vector.
+    Args:
+        anchors: Array of anchors of shape (number of anchors, detections, 2)
+        vector: Vector to calculate cross product with
+
+    Returns:
+        Array of cross products of shape (number of anchors, detections)
+    """
+    vector_at_zero = np.array(
+        [vector.end.x - vector.start.x, vector.end.y - vector.start.y]
+    )
+    vector_start = np.array([vector.start.x, vector.start.y])
+    return np.cross(vector_at_zero, anchors - vector_start)

supervision/utils/internal.py

@@ -1,7 +1,8 @@
 import functools
+import inspect
 import os
 import warnings
-from typing import Callable
+from typing import Any, Callable, Set
 
 
 class SupervisionWarnings(Warning):
@@ -141,3 +142,42 @@ class classproperty(property):
         The result of calling the function stored in 'fget' with 'owner_cls'.
         """
         return self.fget(owner_cls)
+
+
+def get_instance_variables(instance: Any, include_properties=False) -> Set[str]:
+    """
+    Get the public variables of a class instance.
+
+    Args:
+        instance (Any): The instance of a class
+        include_properties (bool): Whether to include properties in the result
+
+    Usage:
+        ```python
+        detections = Detections(xyxy=np.array([1,2,3,4]))
+        variables = get_class_variables(detections)
+        # ["xyxy", "mask", "confidence", ..., "data"]
+        ```
+    """
+    if isinstance(instance, type):
+        raise ValueError("Only class instances are supported, not classes.")
+
+    fields = set(
+        (
+            name
+            for name, val in inspect.getmembers(instance)
+            if not callable(val) and not name.startswith("_")
+        )
+    )
+
+    if not include_properties:
+        properties = set(
+            (
+                name
+                for name, val in inspect.getmembers(instance.__class__)
+                if isinstance(val, property)
+            )
+        )
+        fields -= properties
+
+    return fields

test/dataset/formats/test_coco.py

@@ -1,5 +1,5 @@
 from contextlib import ExitStack as DoesNotRaise
-from typing import Dict, List, Tuple
+from typing import Dict, List, Tuple, Union
 
 import numpy as np
 import pytest
@@ -10,24 +10,30 @@ from supervision.dataset.formats.coco import (
     classes_to_coco_categories,
     coco_annotations_to_detections,
     coco_categories_to_classes,
+    detections_to_coco_annotations,
     group_coco_annotations_by_image_id,
 )
 
 
-def mock_cock_coco_annotation(
+def mock_coco_annotation(
     annotation_id: int = 0,
     image_id: int = 0,
     category_id: int = 0,
     bbox: Tuple[float, float, float, float] = (0.0, 0.0, 0.0, 0.0),
     area: float = 0.0,
+    segmentation: Union[List[list], Dict] = None,
+    iscrowd: bool = False,
 ) -> dict:
+    if not segmentation:
+        segmentation = []
     return {
         "id": annotation_id,
         "image_id": image_id,
         "category_id": category_id,
         "bbox": list(bbox),
         "area": area,
-        "iscrowd": 0,
+        "segmentation": segmentation,
+        "iscrowd": int(iscrowd),
     }
 
 
@@ -101,74 +107,46 @@ def test_classes_to_coco_categories_and_back_to_classes(
     [
         ([], {}, DoesNotRaise()),  # empty coco annotations
         (
-            [mock_cock_coco_annotation(annotation_id=0, image_id=0, category_id=0)],
-            {
-                0: [
-                    mock_cock_coco_annotation(
-                        annotation_id=0, image_id=0, category_id=0
-                    )
-                ]
-            },
+            [mock_coco_annotation(annotation_id=0, image_id=0, category_id=0)],
+            {0: [mock_coco_annotation(annotation_id=0, image_id=0, category_id=0)]},
             DoesNotRaise(),
         ),  # single coco annotation
         (
             [
-                mock_cock_coco_annotation(annotation_id=0, image_id=0, category_id=0),
-                mock_cock_coco_annotation(annotation_id=1, image_id=1, category_id=0),
+                mock_coco_annotation(annotation_id=0, image_id=0, category_id=0),
+                mock_coco_annotation(annotation_id=1, image_id=1, category_id=0),
             ],
             {
-                0: [
-                    mock_cock_coco_annotation(
-                        annotation_id=0, image_id=0, category_id=0
-                    )
-                ],
-                1: [
-                    mock_cock_coco_annotation(
-                        annotation_id=1, image_id=1, category_id=0
-                    )
-                ],
+                0: [mock_coco_annotation(annotation_id=0, image_id=0, category_id=0)],
+                1: [mock_coco_annotation(annotation_id=1, image_id=1, category_id=0)],
             },
             DoesNotRaise(),
         ),  # two coco annotations
         (
             [
-                mock_cock_coco_annotation(annotation_id=0, image_id=0, category_id=0),
-                mock_cock_coco_annotation(annotation_id=1, image_id=1, category_id=1),
-                mock_cock_coco_annotation(annotation_id=2, image_id=1, category_id=2),
-                mock_cock_coco_annotation(annotation_id=3, image_id=2, category_id=3),
-                mock_cock_coco_annotation(annotation_id=4, image_id=3, category_id=1),
-                mock_cock_coco_annotation(annotation_id=5, image_id=3, category_id=2),
-                mock_cock_coco_annotation(annotation_id=5, image_id=3, category_id=3),
+                mock_coco_annotation(annotation_id=0, image_id=0, category_id=0),
+                mock_coco_annotation(annotation_id=1, image_id=1, category_id=1),
+                mock_coco_annotation(annotation_id=2, image_id=1, category_id=2),
+                mock_coco_annotation(annotation_id=3, image_id=2, category_id=3),
+                mock_coco_annotation(annotation_id=4, image_id=3, category_id=1),
+                mock_coco_annotation(annotation_id=5, image_id=3, category_id=2),
+                mock_coco_annotation(annotation_id=5, image_id=3, category_id=3),
             ],
             {
                 0: [
-                    mock_cock_coco_annotation(
-                        annotation_id=0, image_id=0, category_id=0
-                    ),
+                    mock_coco_annotation(annotation_id=0, image_id=0, category_id=0),
                 ],
                 1: [
-                    mock_cock_coco_annotation(
-                        annotation_id=1, image_id=1, category_id=1
-                    ),
-                    mock_cock_coco_annotation(
-                        annotation_id=2, image_id=1, category_id=2
-                    ),
+                    mock_coco_annotation(annotation_id=1, image_id=1, category_id=1),
+                    mock_coco_annotation(annotation_id=2, image_id=1, category_id=2),
                 ],
                 2: [
-                    mock_cock_coco_annotation(
-                        annotation_id=3, image_id=2, category_id=3
-                    ),
+                    mock_coco_annotation(annotation_id=3, image_id=2, category_id=3),
                 ],
                 3: [
-                    mock_cock_coco_annotation(
-                        annotation_id=4, image_id=3, category_id=1
-                    ),
-                    mock_cock_coco_annotation(
-                        annotation_id=5, image_id=3, category_id=2
-                    ),
-                    mock_cock_coco_annotation(
-                        annotation_id=5, image_id=3, category_id=3
-                    ),
+                    mock_coco_annotation(annotation_id=4, image_id=3, category_id=1),
+                    mock_coco_annotation(annotation_id=5, image_id=3, category_id=2),
+                    mock_coco_annotation(annotation_id=5, image_id=3, category_id=3),
                 ],
             },
             DoesNotRaise(),
@@ -195,7 +173,7 @@ def test_group_coco_annotations_by_image_id(
         ),  # empty image annotations
         (
             [
-                mock_cock_coco_annotation(
+                mock_coco_annotation(
                     category_id=0, bbox=(0, 0, 100, 100), area=100 * 100
                 )
             ],
@@ -209,10 +187,10 @@ def test_group_coco_annotations_by_image_id(
         ),  # single image annotations
         (
             [
-                mock_cock_coco_annotation(
+                mock_coco_annotation(
                     category_id=0, bbox=(0, 0, 100, 100), area=100 * 100
                 ),
-                mock_cock_coco_annotation(
+                mock_coco_annotation(
                     category_id=0, bbox=(100, 100, 100, 100), area=100 * 100
                 ),
             ],
@@ -226,6 +204,156 @@ def test_group_coco_annotations_by_image_id(
             ),
             DoesNotRaise(),
         ),  # two image annotations
+        (
+            [
+                mock_coco_annotation(
+                    category_id=0,
+                    bbox=(0, 0, 5, 5),
+                    area=5 * 5,
+                    segmentation=[[0, 0, 2, 0, 2, 2, 4, 2, 4, 4, 0, 4]],
+                )
+            ],
+            (5, 5),
+            True,
+            Detections(
+                xyxy=np.array([[0, 0, 5, 5]], dtype=np.float32),
+                class_id=np.array([0], dtype=int),
+                mask=np.array(
+                    [
+                        [
+                            [1, 1, 1, 0, 0],
+                            [1, 1, 1, 0, 0],
+                            [1, 1, 1, 1, 1],
+                            [1, 1, 1, 1, 1],
+                            [1, 1, 1, 1, 1],
+                        ]
+                    ]
+                ),
+            ),
+            DoesNotRaise(),
+        ),  # single image annotations with mask as polygon
+        (
+            [
+                mock_coco_annotation(
+                    category_id=0,
+                    bbox=(0, 0, 5, 5),
+                    area=5 * 5,
+                    segmentation={
+                        "size": [5, 5],
+                        "counts": [0, 15, 2, 3, 2, 3],
+                    },
+                    iscrowd=True,
+                )
+            ],
+            (5, 5),
+            True,
+            Detections(
+                xyxy=np.array([[0, 0, 5, 5]], dtype=np.float32),
+                class_id=np.array([0], dtype=int),
+                mask=np.array(
+                    [
+                        [
+                            [1, 1, 1, 0, 0],
+                            [1, 1, 1, 0, 0],
+                            [1, 1, 1, 1, 1],
+                            [1, 1, 1, 1, 1],
+                            [1, 1, 1, 1, 1],
+                        ]
+                    ]
+                ),
+            ),
+            DoesNotRaise(),
+        ),  # single image annotations with mask, RLE segmentation mask
+        (
+            [
+                mock_coco_annotation(
+                    category_id=0,
+                    bbox=(0, 0, 5, 5),
+                    area=5 * 5,
+                    segmentation=[[0, 0, 2, 0, 2, 2, 4, 2, 4, 4, 0, 4]],
+                ),
+                mock_coco_annotation(
+                    category_id=0,
+                    bbox=(3, 0, 2, 2),
+                    area=2 * 2,
+                    segmentation={
+                        "size": [5, 5],
+                        "counts": [15, 2, 3, 2, 3],
+                    },
+                    iscrowd=True,
+                ),
+            ],
+            (5, 5),
+            True,
+            Detections(
+                xyxy=np.array([[0, 0, 5, 5], [3, 0, 5, 2]], dtype=np.float32),
+                class_id=np.array([0, 0], dtype=int),
+                mask=np.array(
+                    [
+                        [
+                            [1, 1, 1, 0, 0],
+                            [1, 1, 1, 0, 0],
+                            [1, 1, 1, 1, 1],
+                            [1, 1, 1, 1, 1],
+                            [1, 1, 1, 1, 1],
+                        ],
+                        [
+                            [0, 0, 0, 1, 1],
+                            [0, 0, 0, 1, 1],
+                            [0, 0, 0, 0, 0],
+                            [0, 0, 0, 0, 0],
+                            [0, 0, 0, 0, 0],
+                        ],
+                    ]
+                ),
+            ),
+            DoesNotRaise(),
+        ),  # two image annotations with mask, one mask as polygon ans second as RLE
+        (
+            [
+                mock_coco_annotation(
+                    category_id=0,
+                    bbox=(3, 0, 2, 2),
+                    area=2 * 2,
+                    segmentation={
+                        "size": [5, 5],
+                        "counts": [15, 2, 3, 2, 3],
+                    },
+                    iscrowd=True,
+                ),
+                mock_coco_annotation(
+                    category_id=1,
+                    bbox=(0, 0, 5, 5),
+                    area=5 * 5,
+                    segmentation=[[0, 0, 2, 0, 2, 2, 4, 2, 4, 4, 0, 4]],
+                ),
+            ],
+            (5, 5),
+            True,
+            Detections(
+                xyxy=np.array([[3, 0, 5, 2], [0, 0, 5, 5]], dtype=np.float32),
+                class_id=np.array([0, 1], dtype=int),
+                mask=np.array(
+                    [
+                        [
+                            [0, 0, 0, 1, 1],
+                            [0, 0, 0, 1, 1],
+                            [0, 0, 0, 0, 0],
+                            [0, 0, 0, 0, 0],
+                            [0, 0, 0, 0, 0],
+                        ],
+                        [
+                            [1, 1, 1, 0, 0],
+                            [1, 1, 1, 0, 0],
+                            [1, 1, 1, 1, 1],
+                            [1, 1, 1, 1, 1],
+                            [1, 1, 1, 1, 1],
+                        ],
+                    ]
+                ),
+            ),
+            DoesNotRaise(),
+        ),  # two image annotations with mask, first mask as RLE and second as polygon
     ],
 )
 def test_coco_annotations_to_detections(
@@ -301,3 +429,131 @@ def test_build_coco_class_index_mapping(
             coco_categories=coco_categories, target_classes=target_classes
         )
         assert result == expected_result
+
+
+@pytest.mark.parametrize(
+    "detections, image_id, annotation_id, expected_result, exception",
+    [
+        (
+            Detections(
+                xyxy=np.array([[0, 0, 100, 100]], dtype=np.float32),
+                class_id=np.array([0], dtype=int),
+            ),
+            0,
+            0,
+            [
+                mock_coco_annotation(
+                    category_id=0, bbox=(0, 0, 100, 100), area=100 * 100
+                )
+            ],
+            DoesNotRaise(),
+        ),  # no segmentation mask
+        (
+            Detections(
+                xyxy=np.array([[0, 0, 4, 5]], dtype=np.float32),
+                class_id=np.array([0], dtype=int),
+                mask=np.array(
+                    [
+                        [
+                            [1, 1, 1, 1, 0],
+                            [1, 1, 1, 1, 0],
+                            [1, 1, 1, 1, 0],
+                            [1, 1, 1, 1, 0],
+                            [1, 1, 1, 1, 0],
+                        ]
+                    ]
+                ),
+            ),
+            0,
+            0,
+            [
+                mock_coco_annotation(
+                    category_id=0,
+                    bbox=(0, 0, 4, 5),
+                    area=4 * 5,
+                    segmentation=[[0, 0, 0, 4, 3, 4, 3, 0]],
+                )
+            ],
+            DoesNotRaise(),
+        ),  # segmentation mask in single component,no holes in mask,
+        # expects polygon mask
+        (
+            Detections(
+                xyxy=np.array([[0, 0, 5, 5]], dtype=np.float32),
+                class_id=np.array([0], dtype=int),
+                mask=np.array(
+                    [
+                        [
+                            [1, 1, 1, 0, 0],
+                            [1, 1, 1, 0, 0],
+                            [1, 1, 1, 0, 0],
+                            [0, 0, 0, 1, 1],
+                            [0, 0, 0, 1, 1],
+                        ]
+                    ]
+                ),
+            ),
+            0,
+            0,
+            [
+                mock_coco_annotation(
+                    category_id=0,
+                    bbox=(0, 0, 5, 5),
+                    area=5 * 5,
+                    segmentation={
+                        "size": [5, 5],
+                        "counts": [0, 3, 2, 3, 2, 3, 5, 2, 3, 2],
+                    },
+                    iscrowd=True,
+                )
+            ],
+            DoesNotRaise(),
+        ),  # segmentation mask with 2 components, no holes in mask, expects RLE mask
+        (
+            Detections(
+                xyxy=np.array([[0, 0, 5, 5]], dtype=np.float32),
+                class_id=np.array([0], dtype=int),
+                mask=np.array(
+                    [
+                        [
+                            [0, 1, 1, 1, 1],
+                            [0, 1, 1, 1, 1],
+                            [1, 1, 0, 0, 1],
+                            [1, 1, 0, 0, 1],
+                            [1, 1, 1, 1, 1],
+                        ]
+                    ]
+                ),
+            ),
+            0,
+            0,
+            [
+                mock_coco_annotation(
+                    category_id=0,
+                    bbox=(0, 0, 5, 5),
+                    area=5 * 5,
+                    segmentation={
+                        "size": [5, 5],
+                        "counts": [2, 10, 2, 3, 2, 6],
+                    },
+                    iscrowd=True,
+                )
+            ],
+            DoesNotRaise(),
+        ),  # seg mask in single component, with holes in mask, expects RLE mask
+    ],
+)
+def test_detections_to_coco_annotations(
+    detections: Detections,
+    image_id: int,
+    annotation_id: int,
+    expected_result: List[Dict],
+    exception: Exception,
+) -> None:
+    with exception:
+        result, _ = detections_to_coco_annotations(
+            detections=detections,
+            image_id=image_id,
+            annotation_id=annotation_id,
+        )
+        assert result == expected_result

test/dataset/test_utils.py

@@ -2,13 +2,17 @@ from contextlib import ExitStack as DoesNotRaise
 from test.test_utils import mock_detections
 from typing import Dict, List, Optional, Tuple, TypeVar
 
+import numpy as np
+import numpy.typing as npt
 import pytest
 
 from supervision import Detections
 from supervision.dataset.utils import (
     build_class_index_mapping,
     map_detections_class_id,
+    mask_to_rle,
     merge_class_lists,
+    rle_to_mask,
     train_test_split,
 )
 
@@ -229,3 +233,131 @@ def test_map_detections_class_id(
             source_to_target_mapping=source_to_target_mapping, detections=detections
         )
         assert result == expected_result
+
+
+@pytest.mark.parametrize(
+    "mask, expected_rle, exception",
+    [
+        (
+            np.zeros((3, 3)).astype(bool),
+            [9],
+            DoesNotRaise(),
+        ),  # mask with background only (mask with only False values)
+        (
+            np.ones((3, 3)).astype(bool),
+            [0, 9],
+            DoesNotRaise(),
+        ),  # mask with foreground only (mask with only True values)
+        (
+            np.array(
+                [
+                    [0, 0, 0, 0, 0],
+                    [0, 1, 1, 1, 0],
+                    [0, 1, 0, 1, 0],
+                    [0, 1, 1, 1, 0],
+                    [0, 0, 0, 0, 0],
+                ]
+            ).astype(bool),
+            [6, 3, 2, 1, 1, 1, 2, 3, 6],
+            DoesNotRaise(),
+        ),  # mask where foreground object has hole
+        (
+            np.array(
+                [
+                    [1, 0, 1, 0, 1],
+                    [1, 0, 1, 0, 1],
+                    [1, 0, 1, 0, 1],
+                    [1, 0, 1, 0, 1],
+                    [1, 0, 1, 0, 1],
+                ]
+            ).astype(bool),
+            [0, 5, 5, 5, 5, 5],
+            DoesNotRaise(),
+        ),  # mask where foreground consists of 3 separate components
+        (
+            np.array([[[]]]).astype(bool),
+            None,
+            pytest.raises(AssertionError),
+        ),  # raises AssertionError because mask dimentionality is not 2D
+        (
+            np.array([[]]).astype(bool),
+            None,
+            pytest.raises(AssertionError),
+        ),  # raises AssertionError because mask is empty
+    ],
+)
+def test_mask_to_rle(
+    mask: npt.NDArray[np.bool_], expected_rle: List[int], exception: Exception
+) -> None:
+    with exception:
+        result = mask_to_rle(mask=mask)
+        assert result == expected_rle
+
+
+@pytest.mark.parametrize(
+    "rle, resolution_wh, expected_mask, exception",
+    [
+        (
+            np.array([9]),
+            [3, 3],
+            np.zeros((3, 3)).astype(bool),
+            DoesNotRaise(),
+        ),  # mask with background only (mask with only False values); rle as array
+        (
+            [9],
+            [3, 3],
+            np.zeros((3, 3)).astype(bool),
+            DoesNotRaise(),
+        ),  # mask with background only (mask with only False values); rle as list
+        (
+            np.array([0, 9]),
+            [3, 3],
+            np.ones((3, 3)).astype(bool),
+            DoesNotRaise(),
+        ),  # mask with foreground only (mask with only True values)
+        (
+            np.array([6, 3, 2, 1, 1, 1, 2, 3, 6]),
+            [5, 5],
+            np.array(
+                [
+                    [0, 0, 0, 0, 0],
+                    [0, 1, 1, 1, 0],
+                    [0, 1, 0, 1, 0],
+                    [0, 1, 1, 1, 0],
+                    [0, 0, 0, 0, 0],
+                ]
+            ).astype(bool),
+            DoesNotRaise(),
+        ),  # mask where foreground object has hole
+        (
+            np.array([0, 5, 5, 5, 5, 5]),
+            [5, 5],
+            np.array(
+                [
+                    [1, 0, 1, 0, 1],
+                    [1, 0, 1, 0, 1],
+                    [1, 0, 1, 0, 1],
+                    [1, 0, 1, 0, 1],
+                    [1, 0, 1, 0, 1],
+                ]
+            ).astype(bool),
+            DoesNotRaise(),
+        ),  # mask where foreground consists of 3 separate components
+        (
+            np.array([0, 5, 5, 5, 5, 5]),
+            [2, 2],
+            None,
+            pytest.raises(AssertionError),
+        ),  # raises AssertionError because number of pixels in RLE does not match
+        # number of pixels in expected mask (width x height).
+    ],
+)
+def test_rle_to_mask(
+    rle: npt.NDArray[np.int_],
+    resolution_wh: Tuple[int, int],
+    expected_mask: npt.NDArray[np.bool_],
+    exception: Exception,
+) -> None:
+    with exception:
+        result = rle_to_mask(rle=rle, resolution_wh=resolution_wh)
+        assert np.all(result == expected_mask)

test/detection/test_core.py

@@ -5,7 +5,7 @@ from typing import List, Optional, Union
 import numpy as np
 import pytest
 
-from supervision.detection.core import Detections
+from supervision.detection.core import Detections, merge_inner_detection_object_pair
 from supervision.geometry.core import Position
 
 PREDICTIONS = np.array(
@@ -245,7 +245,6 @@ def test_getitem(
             TEST_DET_1_2,
             DoesNotRaise(),
         ),  # Fields with same keys
-        # Fields and empty
         (
             [TEST_DET_1, Detections.empty()],
             TEST_DET_1,
@@ -264,9 +263,9 @@ def test_getitem(
                 TEST_DET_1,
                 TEST_DET_NONE,
             ],
-            TEST_DET_1,
-            DoesNotRaise(),
-        ),  # Single detection and None fields (+ missing Dict keys)
+            None,
+            pytest.raises(ValueError),
+        ),  # Empty detection, but not Detections.empty()
         # Errors: Non-zero-length differently defined keys & data
         (
             [TEST_DET_1, TEST_DET_DIFFERENT_FIELDS],
@@ -278,6 +277,22 @@ def test_getitem(
             None,
             pytest.raises(ValueError),
         ),  # Non-empty detections with different data keys
+        (
+            [
+                mock_detections(
+                    xyxy=[[10, 10, 20, 20]],
+                    class_id=[1],
+                    mask=[np.zeros((4, 4), dtype=bool)],
+                ),
+                Detections.empty(),
+            ],
+            mock_detections(
+                xyxy=[[10, 10, 20, 20]],
+                class_id=[1],
+                mask=[np.zeros((4, 4), dtype=bool)],
+            ),
+            DoesNotRaise(),
+        ),  # Segmentation + Empty
     ],
 )
 def test_merge(
@@ -421,3 +436,172 @@ def test_equal(
     detections_a: Detections, detections_b: Detections, expected_result: bool
 ) -> None:
     assert (detections_a == detections_b) == expected_result
+
+
+@pytest.mark.parametrize(
+    "detection_1, detection_2, expected_result, exception",
+    [
+        (
+            mock_detections(
+                xyxy=[[10, 10, 30, 30]],
+            ),
+            mock_detections(
+                xyxy=[[10, 10, 30, 30]],
+            ),
+            mock_detections(
+                xyxy=[[10, 10, 30, 30]],
+            ),
+            DoesNotRaise(),
+        ),  # Merge with self
+        (
+            mock_detections(
+                xyxy=[[10, 10, 30, 30]],
+            ),
+            Detections.empty(),
+            None,
+            pytest.raises(ValueError),
+        ),  # merge with empty: error
+        (
+            mock_detections(
+                xyxy=[[10, 10, 30, 30]],
+            ),
+            mock_detections(
+                xyxy=[[10, 10, 30, 30], [40, 40, 60, 60]],
+            ),
+            None,
+            pytest.raises(ValueError),
+        ),  # merge with 2+ objects: error
+        (
+            mock_detections(
+                xyxy=[[10, 10, 30, 30]],
+                confidence=[0.1],
+                class_id=[1],
+                mask=[np.array([[1, 1, 0], [1, 1, 0], [0, 0, 0]], dtype=bool)],
+                tracker_id=[1],
+                data={"key_1": [1]},
+            ),
+            mock_detections(
+                xyxy=[[20, 20, 40, 40]],
+                confidence=[0.1],
+                class_id=[2],
+                mask=[np.array([[0, 0, 0], [0, 1, 1], [0, 1, 1]], dtype=bool)],
+                tracker_id=[2],
+                data={"key_2": [2]},
+            ),
+            mock_detections(
+                xyxy=[[10, 10, 40, 40]],
+                confidence=[0.1],
+                class_id=[1],
+                mask=[np.array([[1, 1, 0], [1, 1, 1], [0, 1, 1]], dtype=bool)],
+                tracker_id=[1],
+                data={"key_1": [1]},
+            ),
+            DoesNotRaise(),
+        ),  # Same confidence - merge box & mask, tie-break to detection_1
+        (
+            mock_detections(
+                xyxy=[[0, 0, 20, 20]],
+                confidence=[0.1],
+                class_id=[1],
+                mask=[np.array([[1, 1, 0], [1, 1, 0], [0, 0, 0]], dtype=bool)],
+                tracker_id=[1],
+                data={"key_1": [1]},
+            ),
+            mock_detections(
+                xyxy=[[10, 10, 50, 50]],
+                confidence=[0.2],
+                class_id=[2],
+                mask=[np.array([[0, 0, 0], [0, 1, 1], [0, 1, 1]], dtype=bool)],
+                tracker_id=[2],
+                data={"key_2": [2]},
+            ),
+            mock_detections(
+                xyxy=[[0, 0, 50, 50]],
+                confidence=[(1 * 0.1 + 4 * 0.2) / 5],
+                class_id=[2],
+                mask=[np.array([[1, 1, 0], [1, 1, 1], [0, 1, 1]], dtype=bool)],
+                tracker_id=[2],
+                data={"key_2": [2]},
+            ),
+            DoesNotRaise(),
+        ),  # Different confidence, different area
+        (
+            mock_detections(
+                xyxy=[[10, 10, 30, 30]],
+                confidence=None,
+                class_id=[1],
+                mask=[np.array([[1, 1, 0], [1, 1, 0], [0, 0, 0]], dtype=bool)],
+                tracker_id=[1],
+                data={"key_1": [1]},
+            ),
+            mock_detections(
+                xyxy=[[20, 20, 40, 40]],
+                confidence=None,
+                class_id=[2],
+                mask=[np.array([[0, 0, 0], [0, 1, 1], [0, 1, 1]], dtype=bool)],
+                tracker_id=[2],
+                data={"key_2": [2]},
+            ),
+            mock_detections(
+                xyxy=[[10, 10, 40, 40]],
+                confidence=None,
+                class_id=[1],
+                mask=[np.array([[1, 1, 0], [1, 1, 1], [0, 1, 1]], dtype=bool)],
+                tracker_id=[1],
+                data={"key_1": [1]},
+            ),
+            DoesNotRaise(),
+        ),  # No confidence at all
+        (
+            mock_detections(
+                xyxy=[[0, 0, 20, 20]],
+                confidence=None,
+            ),
+            mock_detections(
+                xyxy=[[10, 10, 30, 30]],
+                confidence=[0.2],
+            ),
+            None,
+            pytest.raises(ValueError),
+        ),  # confidence: None + [x]
+        (
+            mock_detections(
+                xyxy=[[0, 0, 20, 20]],
+                mask=[np.array([[1, 1, 0], [1, 1, 0], [0, 0, 0]], dtype=bool)],
+            ),
+            mock_detections(
+                xyxy=[[10, 10, 30, 30]],
+                mask=None,
+            ),
+            None,
+            pytest.raises(ValueError),
+        ),  # mask: None + [x]
+        (
+            mock_detections(xyxy=[[0, 0, 20, 20]], tracker_id=[1]),
+            mock_detections(
+                xyxy=[[10, 10, 30, 30]],
+                tracker_id=None,
+            ),
+            None,
+            pytest.raises(ValueError),
+        ),  # tracker_id: None + []
+        (
+            mock_detections(xyxy=[[0, 0, 20, 20]], class_id=[1]),
+            mock_detections(
+                xyxy=[[10, 10, 30, 30]],
+                class_id=None,
+            ),
+            None,
+            pytest.raises(ValueError),
+        ),  # class_id: None + []
+    ],
+)
+def test_merge_inner_detection_object_pair(
+    detection_1: Detections,
+    detection_2: Detections,
+    expected_result: Optional[Detections],
+    exception: Exception,
+):
+    with exception:
+        result = merge_inner_detection_object_pair(detection_1, detection_2)
+        assert result == expected_result

test/detection/test_line_counter.py

@@ -1,10 +1,11 @@
 from contextlib import ExitStack as DoesNotRaise
-from typing import Optional, Tuple
+from test.test_utils import mock_detections
+from typing import List, Optional, Tuple
 
 import pytest
 
 from supervision import LineZone
-from supervision.geometry.core import Point, Vector
+from supervision.geometry.core import Point, Position, Vector
 
 
 @pytest.mark.parametrize(
@@ -70,3 +71,409 @@ def test_calculate_region_of_interest_limits(
     with exception:
         result = LineZone.calculate_region_of_interest_limits(vector=vector)
         assert result == expected_result
+
+
+@pytest.mark.parametrize(
+    "vector, xyxy_sequence, expected_crossed_in, expected_crossed_out",
+    [
+        (  # Vertical line, simple crossing
+            Vector(Point(0, 0), Point(0, 10)),
+            [
+                [4, 4, 6, 6],
+                [4 - 10, 4, 6 - 10, 6],
+                [4, 4, 6, 6],
+                [4 - 10, 4, 6 - 10, 6],
+            ],
+            [False, False, True, False],
+            [False, True, False, True],
+        ),
+        (  # Vertical line reversed, simple crossing
+            Vector(Point(0, 10), Point(0, 0)),
+            [
+                [4, 4, 6, 6],
+                [4 - 10, 4, 6 - 10, 6],
+                [4, 4, 6, 6],
+                [4 - 10, 4, 6 - 10, 6],
+            ],
+            [False, True, False, True],
+            [False, False, True, False],
+        ),
+        (  # Horizontal line, simple crossing
+            Vector(Point(0, 0), Point(10, 0)),
+            [
+                [4, 4, 6, 6],
+                [4, 4 - 10, 6, 6 - 10],
+                [4, 4, 6, 6],
+                [4, 4 - 10, 6, 6 - 10],
+            ],
+            [False, True, False, True],
+            [False, False, True, False],
+        ),
+        (  # Horizontal line reversed, simple crossing
+            Vector(Point(10, 0), Point(0, 0)),
+            [
+                [4, 4, 6, 6],
+                [4, 4 - 10, 6, 6 - 10],
+                [4, 4, 6, 6],
+                [4, 4 - 10, 6, 6 - 10],
+            ],
+            [False, False, True, False],
+            [False, True, False, True],
+        ),
+        (  # Diagonal line, simple crossing
+            Vector(Point(5, 0), Point(0, 5)),
+            [
+                [0, 0, 2, 2],
+                [0 + 10, 0 + 10, 2 + 10, 2 + 10],
+                [0, 0, 2, 2],
+                [0 + 10, 0 + 10, 2 + 10, 2 + 10],
+            ],
+            [False, True, False, True],
+            [False, False, True, False],
+        ),
+        (  # Crossing beside - right side
+            Vector(Point(0, 0), Point(10, 0)),
+            [
+                [20, 4, 24, 6],
+                [20, 4 - 10, 24, 6 - 10],
+                [20, 4, 24, 6],
+                [20, 4 - 10, 24, 6 - 10],
+            ],
+            [False, False, False, False],
+            [False, False, False, False],
+        ),
+        (  # Horizontal line, simple crossing, far away
+            Vector(Point(0, 0), Point(10, 0)),
+            [
+                [4, 1e32, 6, 1e32 + 2],
+                [4, -1e32, 6, -1e32 + 2],
+                [4, 1e32, 6, 1e32 + 2],
+                [4, -1e32, 6, -1e32 + 2],
+            ],
+            [False, True, False, True],
+            [False, False, True, False],
+        ),
+        (  # Crossing beside - left side
+            Vector(Point(0, 0), Point(10, 0)),
+            [
+                [-20, 4, -24, 6],
+                [-20, 4 - 10, -24, 6 - 10],
+                [-20, 4, -24, 6],
+                [-20, 4 - 10, -24, 6 - 10],
+            ],
+            [False, False, False, False],
+            [False, False, False, False],
+        ),
+        (  # Move above
+            Vector(Point(0, 0), Point(10, 0)),
+            [
+                [-4, 4, -2, 6],
+                [-4 + 20, 4, -2 + 20, 6],
+                [-4, 4, -2, 6],
+                [-4 + 20, 4, -2 + 20, 6],
+            ],
+            [False, False, False, False],
+            [False, False, False, False],
+        ),
+        (  # Move below
+            Vector(Point(0, 0), Point(10, 0)),
+            [
+                [-4, -6, -2, -4],
+                [-4 + 20, -6, -2 + 20, -4],
+                [-4, -6, -2, -4],
+                [-4 + 20, -6, -2 + 20, -4],
+            ],
+            [False, False, False, False],
+            [False, False, False, False],
+        ),
+        (  # Move into line partway
+            Vector(Point(0, 0), Point(10, 0)),
+            [
+                [4, 4, 6, 6],
+                [4 + 5, 4, 6 + 5, 6],
+                [4, 4, 6, 6],
+                [4 + 5, 4, 6 + 5, 6],
+            ],
+            [False, False, False, False],
+            [False, False, False, False],
+        ),
+        (  # V-shaped crossing from outside limits - not supported.
+            Vector(Point(0, 0), Point(10, 0)),
+            [[-3, 6, -1, 8], [4, -6, 6, -4], [11, 6, 13, 8]],
+            [False, False, False],
+            [False, False, False],
+        ),
+        (  # Diagonal movement, from within limits to outside - not supported
+            Vector(Point(0, 0), Point(10, 0)),
+            [[4, 1, 6, 3], [11, 1 - 20, 13, 3 - 20]],
+            [False, False],
+            [False, False],
+        ),
+        (  # Diagonal movement, from outside limits to within - not supported
+            Vector(Point(0, 0), Point(10, 0)),
+            [
+                [11, 21, 13, 23],
+                [4, -3, 6, -1],
+            ],
+            [False, False],
+            [False, False],
+        ),
+        (  # Diagonal crossing, from outside to outside limits - not supported.
+            Vector(Point(0, 0), Point(10, 0)),
+            [
+                [-4, 4, -2, 8],
+                [-4 + 16, -4, -2 + 16, -6],
+                [-4, 4, -2, 8],
+                [-4 + 16, -4, -2 + 16, -6],
+            ],
+            [False, False, False, False],
+            [False, False, False, False],
+        ),
+    ],
+)
+def test_line_zone_one_detection_default_anchors(
+    vector: Vector,
+    xyxy_sequence: List[List[float]],
+    expected_crossed_in: List[bool],
+    expected_crossed_out: List[bool],
+) -> None:
+    line_zone = LineZone(start=vector.start, end=vector.end)
+
+    crossed_in_list = []
+    crossed_out_list = []
+    for i, bbox in enumerate(xyxy_sequence):
+        detections = mock_detections(
+            xyxy=[bbox],
+            tracker_id=[0],
+        )
+        crossed_in, crossed_out = line_zone.trigger(detections)
+        crossed_in_list.append(crossed_in[0])
+        crossed_out_list.append(crossed_out[0])
+
+    assert (
+        crossed_in_list == expected_crossed_in
+    ), f"expected {expected_crossed_in}, got {crossed_in_list}"
+    assert (
+        crossed_out_list == expected_crossed_out
+    ), f"expected {expected_crossed_out}, got {crossed_out_list}"
+
+
+@pytest.mark.parametrize(
+    "vector, xyxy_sequence, triggering_anchors, expected_crossed_in, "
+    "expected_crossed_out",
+    [
+        (  # Scrape line, left side, corner anchors
+            Vector(Point(0, 0), Point(10, 0)),
+            [
+                [-2, 4, 2, 6],
+                [-2, 4 - 10, 2, 6 - 10],
+                [-2, 4, 2, 6],
+                [-2, 4 - 10, 2, 6 - 10],
+            ],
+            [
+                Position.TOP_LEFT,
+                Position.BOTTOM_LEFT,
+                Position.TOP_RIGHT,
+                Position.BOTTOM_RIGHT,
+            ],
+            [False, False, False, False],
+            [False, False, False, False],
+        ),
+        (  # Scrape line, left side, right anchors
+            Vector(Point(0, 0), Point(10, 0)),
+            [
+                [-2, 4, 2, 6],
+                [-2, 4 - 10, 2, 6 - 10],
+                [-2, 4, 2, 6],
+                [-2, 4 - 10, 2, 6 - 10],
+            ],
+            [Position.TOP_RIGHT, Position.BOTTOM_RIGHT],
+            [False, True, False, True],
+            [False, False, True, False],
+        ),
+        (  # Scrape line, left side, center anchor (along line point)
+            Vector(Point(0, 0), Point(10, 0)),
+            [
+                [-2, 4, 2, 6],
+                [-2, 4 - 10, 2, 6 - 10],
+                [-2, 4, 2, 6],
+                [-2, 4 - 10, 2, 6 - 10],
+            ],
+            [Position.CENTER],
+            [False, True, False, True],
+            [False, False, True, False],
+        ),
+        (  # Scrape line, right side, corner anchors
+            Vector(Point(0, 0), Point(10, 0)),
+            [
+                [8, 4, 12, 6],
+                [8, 4 - 10, 12, 6 - 10],
+                [8, 4, 12, 6],
+                [8, 4 - 10, 12, 6 - 10],
+            ],
+            [
+                Position.TOP_LEFT,
+                Position.BOTTOM_LEFT,
+                Position.TOP_RIGHT,
+                Position.BOTTOM_RIGHT,
+            ],
+            [False, False, False, False],
+            [False, False, False, False],
+        ),
+        (  # Scrape line, right side, left anchors
+            Vector(Point(0, 0), Point(10, 0)),
+            [
+                [8, 4, 12, 6],
+                [8, 4 - 10, 12, 6 - 10],
+                [8, 4, 12, 6],
+                [8, 4 - 10, 12, 6 - 10],
+            ],
+            [Position.TOP_LEFT, Position.BOTTOM_LEFT],
+            [False, True, False, True],
+            [False, False, True, False],
+        ),
+        (  # Scrape line, right side, center anchor (along line point)
+            Vector(Point(0, 0), Point(10, 0)),
+            [
+                [8, 4, 12, 6],
+                [8, 4 - 10, 12, 6 - 10],
+                [8, 4, 12, 6],
+                [8, 4 - 10, 12, 6 - 10],
+            ],
+            [Position.CENTER],
+            [False, True, False, True],
+            [False, False, True, False],
+        ),
+        (  # Simple crossing, one anchor
+            Vector(Point(0, 0), Point(10, 0)),
+            [
+                [4, 4, 6, 6],
+                [4, 4 - 10, 6, 6 - 10],
+                [4, 4, 6, 6],
+                [4, 4 - 10, 6, 6 - 10],
+            ],
+            [Position.CENTER],
+            [False, True, False, True],
+            [False, False, True, False],
+        ),
+        (  # Simple crossing, all box anchors
+            Vector(Point(0, 0), Point(10, 0)),
+            [
+                [4, 4, 6, 6],
+                [4, 4 - 10, 6, 6 - 10],
+                [4, 4, 6, 6],
+                [4, 4 - 10, 6, 6 - 10],
+            ],
+            [
+                Position.CENTER,
+                Position.CENTER_LEFT,
+                Position.CENTER_RIGHT,
+                Position.TOP_CENTER,
+                Position.TOP_LEFT,
+                Position.TOP_RIGHT,
+                Position.BOTTOM_LEFT,
+                Position.BOTTOM_CENTER,
+                Position.BOTTOM_RIGHT,
+            ],
+            [False, True, False, True],
+            [False, False, True, False],
+        ),
+    ],
+)
+def test_line_zone_one_detection(
+    vector: Vector,
+    xyxy_sequence: List[List[float]],
+    triggering_anchors: List[Position],
+    expected_crossed_in: List[bool],
+    expected_crossed_out: List[bool],
+) -> None:
+    line_zone = LineZone(
+        start=vector.start, end=vector.end, triggering_anchors=triggering_anchors
+    )
+
+    crossed_in_list = []
+    crossed_out_list = []
+    for i, bbox in enumerate(xyxy_sequence):
+        detections = mock_detections(
+            xyxy=[bbox],
+            tracker_id=[0],
+        )
+        crossed_in, crossed_out = line_zone.trigger(detections)
+        crossed_in_list.append(crossed_in[0])
+        crossed_out_list.append(crossed_out[0])
+
+    assert (
+        crossed_in_list == expected_crossed_in
+    ), f"expected {expected_crossed_in}, got {crossed_in_list}"
+    assert (
+        crossed_out_list == expected_crossed_out
+    ), f"expected {expected_crossed_out}, got {crossed_out_list}"
+
+
+@pytest.mark.parametrize(
+    "vector, xyxy_sequence, anchors, expected_crossed_in, "
+    "expected_crossed_out, exception",
+    [
+        (  # One stays, one crosses
+            Vector(Point(0, 0), Point(10, 0)),
+            [
+                [[4, 4, 6, 6], [4, 4, 6, 6]],
+                [[4, 4, 6, 6], [4, 4 - 10, 6, 6 - 10]],
+                [[4, 4, 6, 6], [4, 4, 6, 6]],
+                [[4, 4, 6, 6], [4, 4 - 10, 6, 6 - 10]],
+            ],
+            [
+                Position.TOP_LEFT,
+                Position.TOP_RIGHT,
+                Position.BOTTOM_LEFT,
+                Position.BOTTOM_RIGHT,
+            ],
+            [[False, False], [False, True], [False, False], [False, True]],
+            [[False, False], [False, False], [False, True], [False, False]],
+            DoesNotRaise(),
+        ),
+        (  # Both cross at the same time
+            Vector(Point(0, 0), Point(10, 0)),
+            [
+                [[4, 4, 6, 6], [4, 4, 6, 6]],
+                [[4, 4 - 10, 6, 6 - 10], [4, 4 - 10, 6, 6 - 10]],
+                [[4, 4, 6, 6], [4, 4, 6, 6]],
+                [[4, 4 - 10, 6, 6 - 10], [4, 4 - 10, 6, 6 - 10]],
+            ],
+            [
+                Position.TOP_LEFT,
+                Position.TOP_RIGHT,
+                Position.BOTTOM_LEFT,
+                Position.BOTTOM_RIGHT,
+            ],
+            [[False, False], [True, True], [False, False], [True, True]],
+            [[False, False], [False, False], [True, True], [False, False]],
+            DoesNotRaise(),
+        ),
+    ],
+)
+def test_line_zone_multiple_detections(
+    vector: Vector,
+    xyxy_sequence: List[List[List[float]]],
+    anchors: List[Position],
+    expected_crossed_in: List[List[bool]],
+    expected_crossed_out: List[List[bool]],
+    exception: Exception,
+) -> None:
+    with exception:
+        line_zone = LineZone(
+            start=vector.start, end=vector.end, triggering_anchors=anchors
+        )
+        crossed_in_list = []
+        crossed_out_list = []
+        for bboxes in xyxy_sequence:
+            detections = mock_detections(
+                xyxy=bboxes,
+                tracker_id=[i for i in range(0, len(bboxes))],
+            )
+            crossed_in, crossed_out = line_zone.trigger(detections)
+            crossed_in_list.append(list(crossed_in))
+            crossed_out_list.append(list(crossed_out))
+
+        assert crossed_in_list == expected_crossed_in
+        assert crossed_out_list == expected_crossed_out

test/detection/test_lmm.py

@@ -0,0 +1,151 @@
+from typing import List, Optional, Tuple
+
+import numpy as np
+import pytest
+
+from supervision.detection.lmm import from_paligemma
+
+
+@pytest.mark.parametrize(
+    "result, resolution_wh, classes, expected_results",
+    [
+        (
+            "",
+            (1000, 1000),
+            None,
+            (np.empty((0, 4)), None, np.empty(0).astype(str)),
+        ),  # empty response
+        (
+            "",
+            (1000, 1000),
+            ["cat", "dog"],
+            (np.empty((0, 4)), None, np.empty(0).astype(str)),
+        ),  # empty response with classes
+        (
+            "\n",
+            (1000, 1000),
+            None,
+            (np.empty((0, 4)), None, np.empty(0).astype(str)),
+        ),  # new line response
+        (
+            "the quick brown fox jumps over the lazy dog.",
+            (1000, 1000),
+            None,
+            (np.empty((0, 4)), None, np.empty(0).astype(str)),
+        ),  # response with no location
+        (
+            "<loc0256><loc0768><loc0768> cat",
+            (1000, 1000),
+            None,
+            (np.empty((0, 4)), None, np.empty(0).astype(str)),
+        ),  # response with missing location
+        (
+            "<loc0256><loc0256><loc0768><loc0768><loc0768> cat",
+            (1000, 1000),
+            None,
+            (np.empty((0, 4)), None, np.empty(0).astype(str)),
+        ),  # response with extra location
+        (
+            "<loc0256><loc0256><loc0768><loc0768>",
+            (1000, 1000),
+            None,
+            (np.empty((0, 4)), None, np.empty(0).astype(str)),
+        ),  # response with no class
+        (
+            "<loc0256><loc0256><loc0768><loc0768> catt",
+            (1000, 1000),
+            ["cat", "dog"],
+            (np.empty((0, 4)), np.empty(0), np.empty(0).astype(str)),
+        ),  # response with invalid class
+        (
+            "<loc0256><loc0256><loc0768><loc0768> cat",
+            (1000, 1000),
+            None,
+            (
+                np.array([[250.0, 250.0, 750.0, 750.0]]),
+                None,
+                np.array(["cat"]).astype(str),
+            ),
+        ),  # correct response; no classes
+        (
+            "<loc0256><loc0256><loc0768><loc0768> black cat",
+            (1000, 1000),
+            None,
+            (
+                np.array([[250.0, 250.0, 750.0, 750.0]]),
+                None,
+                np.array(["black cat"]).astype(np.dtype("U")),
+            ),
+        ),  # correct response; class name with space; no classes
+        (
+            "<loc0256><loc0256><loc0768><loc0768> black-cat",
+            (1000, 1000),
+            None,
+            (
+                np.array([[250.0, 250.0, 750.0, 750.0]]),
+                None,
+                np.array(["black-cat"]).astype(np.dtype("U")),
+            ),
+        ),  # correct response; class name with hyphen; no classes
+        (
+            "<loc0256><loc0256><loc0768><loc0768> black_cat",
+            (1000, 1000),
+            None,
+            (
+                np.array([[250.0, 250.0, 750.0, 750.0]]),
+                None,
+                np.array(["black_cat"]).astype(np.dtype("U")),
+            ),
+        ),  # correct response; class name with underscore; no classes
+        (
+            "<loc0256><loc0256><loc0768><loc0768> cat ;",
+            (1000, 1000),
+            ["cat", "dog"],
+            (
+                np.array([[250.0, 250.0, 750.0, 750.0]]),
+                np.array([0]),
+                np.array(["cat"]).astype(str),
+            ),
+        ),  # correct response; with classes
+        (
+            "<loc0256><loc0256><loc0768><loc0768> cat ; <loc0256><loc0256><loc0768><loc0768> dog",  # noqa: E501
+            (1000, 1000),
+            ["cat", "dog"],
+            (
+                np.array([[250.0, 250.0, 750.0, 750.0], [250.0, 250.0, 750.0, 750.0]]),
+                np.array([0, 1]),
+                np.array(["cat", "dog"]).astype(np.dtype("U")),
+            ),
+        ),  # correct response; with classes
+        (
+            "<loc0256><loc0256><loc0768><loc0768> cat ; <loc0256><loc0256><loc0768> cat",  # noqa: E501
+            (1000, 1000),
+            ["cat", "dog"],
+            (
+                np.array([[250.0, 250.0, 750.0, 750.0]]),
+                np.array([0]),
+                np.array(["cat"]).astype(str),
+            ),
+        ),  # partially correct response; with classes
+        (
+            "<loc0256><loc0256><loc0768><loc0768> cat ; <loc0256><loc0256><loc0768><loc0768><loc0768> cat",  # noqa: E501
+            (1000, 1000),
+            ["cat", "dog"],
+            (
+                np.array([[250.0, 250.0, 750.0, 750.0]]),
+                np.array([0]),
+                np.array(["cat"]).astype(str),
+            ),
+        ),  # partially correct response; with classes
+    ],
+)
+def test_from_paligemma(
+    result: str,
+    resolution_wh: Tuple[int, int],
+    classes: Optional[List[str]],
+    expected_results: Tuple[np.ndarray, Optional[np.ndarray], np.ndarray],
+) -> None:
+    result = from_paligemma(result=result, resolution_wh=resolution_wh, classes=classes)
+    np.testing.assert_array_equal(result[0], expected_results[0])
+    np.testing.assert_array_equal(result[1], expected_results[1])
+    np.testing.assert_array_equal(result[2], expected_results[2])

test/detection/test_overlap_filter.py

@@ -0,0 +1,449 @@
+from contextlib import ExitStack as DoesNotRaise
+from typing import List, Optional
+
+import numpy as np
+import pytest
+
+from supervision.detection.overlap_filter import (
+    box_non_max_suppression,
+    group_overlapping_boxes,
+    mask_non_max_suppression,
+)
+
+
+@pytest.mark.parametrize(
+    "predictions, iou_threshold, expected_result, exception",
+    [
+        (
+            np.empty(shape=(0, 5), dtype=float),
+            0.5,
+            [],
+            DoesNotRaise(),
+        ),
+        (
+            np.array([[0, 0, 10, 10, 1.0]]),
+            0.5,
+            [[0]],
+            DoesNotRaise(),
+        ),
+        (
+            np.array([[0, 0, 10, 10, 1.0], [0, 0, 9, 9, 1.0]]),
+            0.5,
+            [[1, 0]],
+            DoesNotRaise(),
+        ),  # High overlap, tie-break to second det
+        (
+            np.array([[0, 0, 10, 10, 1.0], [0, 0, 9, 9, 0.99]]),
+            0.5,
+            [[0, 1]],
+            DoesNotRaise(),
+        ),  # High overlap, merge to high confidence
+        (
+            np.array([[0, 0, 10, 10, 0.99], [0, 0, 9, 9, 1.0]]),
+            0.5,
+            [[1, 0]],
+            DoesNotRaise(),
+        ),  # (test symmetry) High overlap, merge to high confidence
+        (
+            np.array([[0, 0, 10, 10, 0.90], [0, 0, 9, 9, 1.0]]),
+            0.5,
+            [[1, 0]],
+            DoesNotRaise(),
+        ),  # (test symmetry) High overlap, merge to high confidence
+        (
+            np.array([[0, 0, 10, 10, 1.0], [0, 0, 9, 9, 1.0]]),
+            1.0,
+            [[1], [0]],
+            DoesNotRaise(),
+        ),  # High IOU required
+        (
+            np.array([[0, 0, 10, 10, 1.0], [0, 0, 9, 9, 1.0]]),
+            0.0,
+            [[1, 0]],
+            DoesNotRaise(),
+        ),  # No IOU required
+        (
+            np.array([[0, 0, 10, 10, 1.0], [0, 0, 5, 5, 0.9]]),
+            0.25,
+            [[0, 1]],
+            DoesNotRaise(),
+        ),  # Below IOU requirement
+        (
+            np.array([[0, 0, 10, 10, 1.0], [0, 0, 5, 5, 0.9]]),
+            0.26,
+            [[0], [1]],
+            DoesNotRaise(),
+        ),  # Above IOU requirement
+        (
+            np.array([[0, 0, 10, 10, 1.0], [0, 0, 9, 9, 1.0], [0, 0, 8, 8, 1.0]]),
+            0.5,
+            [[2, 1, 0]],
+            DoesNotRaise(),
+        ),  # 3 boxes
+        (
+            np.array(
+                [
+                    [0, 0, 10, 10, 1.0],
+                    [0, 0, 9, 9, 1.0],
+                    [5, 5, 10, 10, 1.0],
+                    [6, 6, 10, 10, 1.0],
+                    [9, 9, 10, 10, 1.0],
+                ]
+            ),
+            0.5,
+            [[4], [3, 2], [1, 0]],
+            DoesNotRaise(),
+        ),  # 5 boxes, 2 merges, 1 separate
+        (
+            np.array(
+                [
+                    [0, 0, 2, 1, 1.0],
+                    [1, 0, 3, 1, 1.0],
+                    [2, 0, 4, 1, 1.0],
+                    [3, 0, 5, 1, 1.0],
+                    [4, 0, 6, 1, 1.0],
+                ]
+            ),
+            0.33,
+            [[4, 3], [2, 1], [0]],
+            DoesNotRaise(),
+        ),  # sequential merge, half overlap
+        (
+            np.array(
+                [
+                    [0, 0, 2, 1, 0.9],
+                    [1, 0, 3, 1, 0.9],
+                    [2, 0, 4, 1, 1.0],
+                    [3, 0, 5, 1, 0.9],
+                    [4, 0, 6, 1, 0.9],
+                ]
+            ),
+            0.33,
+            [[2, 3, 1], [4], [0]],
+            DoesNotRaise(),
+        ),  # confidence
+    ],
+)
+def test_group_overlapping_boxes(
+    predictions: np.ndarray,
+    iou_threshold: float,
+    expected_result: List[List[int]],
+    exception: Exception,
+) -> None:
+    with exception:
+        result = group_overlapping_boxes(
+            predictions=predictions, iou_threshold=iou_threshold
+        )
+
+        assert result == expected_result
+
+
+@pytest.mark.parametrize(
+    "predictions, iou_threshold, expected_result, exception",
+    [
+        (
+            np.empty(shape=(0, 5)),
+            0.5,
+            np.array([]),
+            DoesNotRaise(),
+        ),  # single box with no category
+        (
+            np.array([[10.0, 10.0, 40.0, 40.0, 0.8]]),
+            0.5,
+            np.array([True]),
+            DoesNotRaise(),
+        ),  # single box with no category
+        (
+            np.array([[10.0, 10.0, 40.0, 40.0, 0.8, 0]]),
+            0.5,
+            np.array([True]),
+            DoesNotRaise(),
+        ),  # single box with category
+        (
+            np.array(
+                [
+                    [10.0, 10.0, 40.0, 40.0, 0.8],
+                    [15.0, 15.0, 40.0, 40.0, 0.9],
+                ]
+            ),
+            0.5,
+            np.array([False, True]),
+            DoesNotRaise(),
+        ),  # two boxes with no category
+        (
+            np.array(
+                [
+                    [10.0, 10.0, 40.0, 40.0, 0.8, 0],
+                    [15.0, 15.0, 40.0, 40.0, 0.9, 1],
+                ]
+            ),
+            0.5,
+            np.array([True, True]),
+            DoesNotRaise(),
+        ),  # two boxes with different category
+        (
+            np.array(
+                [
+                    [10.0, 10.0, 40.0, 40.0, 0.8, 0],
+                    [15.0, 15.0, 40.0, 40.0, 0.9, 0],
+                ]
+            ),
+            0.5,
+            np.array([False, True]),
+            DoesNotRaise(),
+        ),  # two boxes with same category
+        (
+            np.array(
+                [
+                    [0.0, 0.0, 30.0, 40.0, 0.8],
+                    [5.0, 5.0, 35.0, 45.0, 0.9],
+                    [10.0, 10.0, 40.0, 50.0, 0.85],
+                ]
+            ),
+            0.5,
+            np.array([False, True, False]),
+            DoesNotRaise(),
+        ),  # three boxes with no category
+        (
+            np.array(
+                [
+                    [0.0, 0.0, 30.0, 40.0, 0.8, 0],
+                    [5.0, 5.0, 35.0, 45.0, 0.9, 1],
+                    [10.0, 10.0, 40.0, 50.0, 0.85, 2],
+                ]
+            ),
+            0.5,
+            np.array([True, True, True]),
+            DoesNotRaise(),
+        ),  # three boxes with same category
+        (
+            np.array(
+                [
+                    [0.0, 0.0, 30.0, 40.0, 0.8, 0],
+                    [5.0, 5.0, 35.0, 45.0, 0.9, 0],
+                    [10.0, 10.0, 40.0, 50.0, 0.85, 1],
+                ]
+            ),
+            0.5,
+            np.array([False, True, True]),
+            DoesNotRaise(),
+        ),  # three boxes with different category
+    ],
+)
+def test_box_non_max_suppression(
+    predictions: np.ndarray,
+    iou_threshold: float,
+    expected_result: Optional[np.ndarray],
+    exception: Exception,
+) -> None:
+    with exception:
+        result = box_non_max_suppression(
+            predictions=predictions, iou_threshold=iou_threshold
+        )
+        assert np.array_equal(result, expected_result)
+
+
+@pytest.mark.parametrize(
+    "predictions, masks, iou_threshold, expected_result, exception",
+    [
+        (
+            np.empty((0, 6)),
+            np.empty((0, 5, 5)),
+            0.5,
+            np.array([]),
+            DoesNotRaise(),
+        ),  # empty predictions and masks
+        (
+            np.array([[0, 0, 0, 0, 0.8]]),
+            np.array(
+                [
+                    [
+                        [False, False, False, False, False],
+                        [False, True, True, True, False],
+                        [False, True, True, True, False],
+                        [False, True, True, True, False],
+                        [False, False, False, False, False],
+                    ]
+                ]
+            ),
+            0.5,
+            np.array([True]),
+            DoesNotRaise(),
+        ),  # single mask with no category
+        (
+            np.array([[0, 0, 0, 0, 0.8, 0]]),
+            np.array(
+                [
+                    [
+                        [False, False, False, False, False],
+                        [False, True, True, True, False],
+                        [False, True, True, True, False],
+                        [False, True, True, True, False],
+                        [False, False, False, False, False],
+                    ]
+                ]
+            ),
+            0.5,
+            np.array([True]),
+            DoesNotRaise(),
+        ),  # single mask with category
+        (
+            np.array([[0, 0, 0, 0, 0.8], [0, 0, 0, 0, 0.9]]),
+            np.array(
+                [
+                    [
+                        [False, False, False, False, False],
+                        [False, True, True, False, False],
+                        [False, True, True, False, False],
+                        [False, False, False, False, False],
+                        [False, False, False, False, False],
+                    ],
+                    [
+                        [False, False, False, False, False],
+                        [False, False, False, False, False],
+                        [False, False, False, True, True],
+                        [False, False, False, True, True],
+                        [False, False, False, False, False],
+                    ],
+                ]
+            ),
+            0.5,
+            np.array([True, True]),
+            DoesNotRaise(),
+        ),  # two masks non-overlapping with no category
+        (
+            np.array([[0, 0, 0, 0, 0.8], [0, 0, 0, 0, 0.9]]),
+            np.array(
+                [
+                    [
+                        [False, False, False, False, False],
+                        [False, True, True, True, False],
+                        [False, True, True, True, False],
+                        [False, True, True, True, False],
+                        [False, False, False, False, False],
+                    ],
+                    [
+                        [False, False, False, False, False],
+                        [False, False, True, True, True],
+                        [False, False, True, True, True],
+                        [False, False, True, True, True],
+                        [False, False, False, False, False],
+                    ],
+                ]
+            ),
+            0.4,
+            np.array([False, True]),
+            DoesNotRaise(),
+        ),  # two masks partially overlapping with no category
+        (
+            np.array([[0, 0, 0, 0, 0.8, 0], [0, 0, 0, 0, 0.9, 1]]),
+            np.array(
+                [
+                    [
+                        [False, False, False, False, False],
+                        [False, True, True, True, False],
+                        [False, True, True, True, False],
+                        [False, True, True, True, False],
+                        [False, False, False, False, False],
+                    ],
+                    [
+                        [False, False, False, False, False],
+                        [False, False, True, True, True],
+                        [False, False, True, True, True],
+                        [False, False, True, True, True],
+                        [False, False, False, False, False],
+                    ],
+                ]
+            ),
+            0.5,
+            np.array([True, True]),
+            DoesNotRaise(),
+        ),  # two masks partially overlapping with different category
+        (
+            np.array(
+                [
+                    [0, 0, 0, 0, 0.8],
+                    [0, 0, 0, 0, 0.85],
+                    [0, 0, 0, 0, 0.9],
+                ]
+            ),
+            np.array(
+                [
+                    [
+                        [False, False, False, False, False],
+                        [False, True, True, False, False],
+                        [False, True, True, False, False],
+                        [False, False, False, False, False],
+                        [False, False, False, False, False],
+                    ],
+                    [
+                        [False, False, False, False, False],
+                        [False, True, True, False, False],
+                        [False, True, True, False, False],
+                        [False, False, False, False, False],
+                        [False, False, False, False, False],
+                    ],
+                    [
+                        [False, False, False, False, False],
+                        [False, False, False, True, True],
+                        [False, False, False, True, True],
+                        [False, False, False, False, False],
+                        [False, False, False, False, False],
+                    ],
+                ]
+            ),
+            0.5,
+            np.array([False, True, True]),
+            DoesNotRaise(),
+        ),  # three masks with no category
+        (
+            np.array(
+                [
+                    [0, 0, 0, 0, 0.8, 0],
+                    [0, 0, 0, 0, 0.85, 1],
+                    [0, 0, 0, 0, 0.9, 2],
+                ]
+            ),
+            np.array(
+                [
+                    [
+                        [False, False, False, False, False],
+                        [False, True, True, False, False],
+                        [False, True, True, False, False],
+                        [False, False, False, False, False],
+                        [False, False, False, False, False],
+                    ],
+                    [
+                        [False, False, False, False, False],
+                        [False, True, True, False, False],
+                        [False, True, True, False, False],
+                        [False, True, True, False, False],
+                        [False, False, False, False, False],
+                    ],
+                    [
+                        [False, False, False, False, False],
+                        [False, True, True, False, False],
+                        [False, True, True, False, False],
+                        [False, False, False, False, False],
+                        [False, False, False, False, False],
+                    ],
+                ]
+            ),
+            0.5,
+            np.array([True, True, True]),
+            DoesNotRaise(),
+        ),  # three masks with different category
+    ],
+)
+def test_mask_non_max_suppression(
+    predictions: np.ndarray,
+    masks: np.ndarray,
+    iou_threshold: float,
+    expected_result: Optional[np.ndarray],
+    exception: Exception,
+) -> None:
+    with exception:
+        result = mask_non_max_suppression(
+            predictions=predictions, masks=masks, iou_threshold=iou_threshold
+        )
+        assert np.array_equal(result, expected_result)

test/detection/test_polygonzone.py

@@ -92,3 +92,19 @@ def test_polygon_zone_trigger(
     with exception:
         in_zone = polygon_zone.trigger(detections)
         assert np.all(in_zone == expected_results)
+
+
+@pytest.mark.parametrize(
+    "polygon, triggering_anchors, exception",
+    [
+        (POLYGON, [sv.Position.CENTER], DoesNotRaise()),
+        (
+            POLYGON,
+            [],
+            pytest.raises(ValueError),
+        ),
+    ],
+)
+def test_polygon_zone_initialization(polygon, triggering_anchors, exception):
+    with exception:
+        sv.PolygonZone(polygon, FRAME_RESOLUTION, triggering_anchors=triggering_anchors)

test/detection/test_utils.py

@@ -2,16 +2,17 @@ from contextlib import ExitStack as DoesNotRaise
 from typing import Any, Dict, List, Optional, Tuple
 
 import numpy as np
+import numpy.typing as npt
 import pytest
 
 from supervision.config import CLASS_NAME_DATA_FIELD
 from supervision.detection.utils import (
-    box_non_max_suppression,
     calculate_masks_centroids,
     clip_boxes,
+    contains_holes,
+    contains_multiple_segments,
     filter_polygons_by_area,
     get_data_item,
-    mask_non_max_suppression,
     merge_data,
     move_boxes,
     process_roboflow_result,
@@ -22,317 +23,6 @@ TEST_MASK = np.zeros((1, 1000, 1000), dtype=bool)
 TEST_MASK[:, 300:351, 200:251] = True
 
 
-@pytest.mark.parametrize(
-    "predictions, iou_threshold, expected_result, exception",
-    [
-        (
-            np.empty(shape=(0, 5)),
-            0.5,
-            np.array([]),
-            DoesNotRaise(),
-        ),  # single box with no category
-        (
-            np.array([[10.0, 10.0, 40.0, 40.0, 0.8]]),
-            0.5,
-            np.array([True]),
-            DoesNotRaise(),
-        ),  # single box with no category
-        (
-            np.array([[10.0, 10.0, 40.0, 40.0, 0.8, 0]]),
-            0.5,
-            np.array([True]),
-            DoesNotRaise(),
-        ),  # single box with category
-        (
-            np.array(
-                [
-                    [10.0, 10.0, 40.0, 40.0, 0.8],
-                    [15.0, 15.0, 40.0, 40.0, 0.9],
-                ]
-            ),
-            0.5,
-            np.array([False, True]),
-            DoesNotRaise(),
-        ),  # two boxes with no category
-        (
-            np.array(
-                [
-                    [10.0, 10.0, 40.0, 40.0, 0.8, 0],
-                    [15.0, 15.0, 40.0, 40.0, 0.9, 1],
-                ]
-            ),
-            0.5,
-            np.array([True, True]),
-            DoesNotRaise(),
-        ),  # two boxes with different category
-        (
-            np.array(
-                [
-                    [10.0, 10.0, 40.0, 40.0, 0.8, 0],
-                    [15.0, 15.0, 40.0, 40.0, 0.9, 0],
-                ]
-            ),
-            0.5,
-            np.array([False, True]),
-            DoesNotRaise(),
-        ),  # two boxes with same category
-        (
-            np.array(
-                [
-                    [0.0, 0.0, 30.0, 40.0, 0.8],
-                    [5.0, 5.0, 35.0, 45.0, 0.9],
-                    [10.0, 10.0, 40.0, 50.0, 0.85],
-                ]
-            ),
-            0.5,
-            np.array([False, True, False]),
-            DoesNotRaise(),
-        ),  # three boxes with no category
-        (
-            np.array(
-                [
-                    [0.0, 0.0, 30.0, 40.0, 0.8, 0],
-                    [5.0, 5.0, 35.0, 45.0, 0.9, 1],
-                    [10.0, 10.0, 40.0, 50.0, 0.85, 2],
-                ]
-            ),
-            0.5,
-            np.array([True, True, True]),
-            DoesNotRaise(),
-        ),  # three boxes with same category
-        (
-            np.array(
-                [
-                    [0.0, 0.0, 30.0, 40.0, 0.8, 0],
-                    [5.0, 5.0, 35.0, 45.0, 0.9, 0],
-                    [10.0, 10.0, 40.0, 50.0, 0.85, 1],
-                ]
-            ),
-            0.5,
-            np.array([False, True, True]),
-            DoesNotRaise(),
-        ),  # three boxes with different category
-    ],
-)
-def test_box_non_max_suppression(
-    predictions: np.ndarray,
-    iou_threshold: float,
-    expected_result: Optional[np.ndarray],
-    exception: Exception,
-) -> None:
-    with exception:
-        result = box_non_max_suppression(
-            predictions=predictions, iou_threshold=iou_threshold
-        )
-        assert np.array_equal(result, expected_result)
-
-
-@pytest.mark.parametrize(
-    "predictions, masks, iou_threshold, expected_result, exception",
-    [
-        (
-            np.empty((0, 6)),
-            np.empty((0, 5, 5)),
-            0.5,
-            np.array([]),
-            DoesNotRaise(),
-        ),  # empty predictions and masks
-        (
-            np.array([[0, 0, 0, 0, 0.8]]),
-            np.array(
-                [
-                    [
-                        [False, False, False, False, False],
-                        [False, True, True, True, False],
-                        [False, True, True, True, False],
-                        [False, True, True, True, False],
-                        [False, False, False, False, False],
-                    ]
-                ]
-            ),
-            0.5,
-            np.array([True]),
-            DoesNotRaise(),
-        ),  # single mask with no category
-        (
-            np.array([[0, 0, 0, 0, 0.8, 0]]),
-            np.array(
-                [
-                    [
-                        [False, False, False, False, False],
-                        [False, True, True, True, False],
-                        [False, True, True, True, False],
-                        [False, True, True, True, False],
-                        [False, False, False, False, False],
-                    ]
-                ]
-            ),
-            0.5,
-            np.array([True]),
-            DoesNotRaise(),
-        ),  # single mask with category
-        (
-            np.array([[0, 0, 0, 0, 0.8], [0, 0, 0, 0, 0.9]]),
-            np.array(
-                [
-                    [
-                        [False, False, False, False, False],
-                        [False, True, True, False, False],
-                        [False, True, True, False, False],
-                        [False, False, False, False, False],
-                        [False, False, False, False, False],
-                    ],
-                    [
-                        [False, False, False, False, False],
-                        [False, False, False, False, False],
-                        [False, False, False, True, True],
-                        [False, False, False, True, True],
-                        [False, False, False, False, False],
-                    ],
-                ]
-            ),
-            0.5,
-            np.array([True, True]),
-            DoesNotRaise(),
-        ),  # two masks non-overlapping with no category
-        (
-            np.array([[0, 0, 0, 0, 0.8], [0, 0, 0, 0, 0.9]]),
-            np.array(
-                [
-                    [
-                        [False, False, False, False, False],
-                        [False, True, True, True, False],
-                        [False, True, True, True, False],
-                        [False, True, True, True, False],
-                        [False, False, False, False, False],
-                    ],
-                    [
-                        [False, False, False, False, False],
-                        [False, False, True, True, True],
-                        [False, False, True, True, True],
-                        [False, False, True, True, True],
-                        [False, False, False, False, False],
-                    ],
-                ]
-            ),
-            0.4,
-            np.array([False, True]),
-            DoesNotRaise(),
-        ),  # two masks partially overlapping with no category
-        (
-            np.array([[0, 0, 0, 0, 0.8, 0], [0, 0, 0, 0, 0.9, 1]]),
-            np.array(
-                [
-                    [
-                        [False, False, False, False, False],
-                        [False, True, True, True, False],
-                        [False, True, True, True, False],
-                        [False, True, True, True, False],
-                        [False, False, False, False, False],
-                    ],
-                    [
-                        [False, False, False, False, False],
-                        [False, False, True, True, True],
-                        [False, False, True, True, True],
-                        [False, False, True, True, True],
-                        [False, False, False, False, False],
-                    ],
-                ]
-            ),
-            0.5,
-            np.array([True, True]),
-            DoesNotRaise(),
-        ),  # two masks partially overlapping with different category
-        (
-            np.array(
-                [
-                    [0, 0, 0, 0, 0.8],
-                    [0, 0, 0, 0, 0.85],
-                    [0, 0, 0, 0, 0.9],
-                ]
-            ),
-            np.array(
-                [
-                    [
-                        [False, False, False, False, False],
-                        [False, True, True, False, False],
-                        [False, True, True, False, False],
-                        [False, False, False, False, False],
-                        [False, False, False, False, False],
-                    ],
-                    [
-                        [False, False, False, False, False],
-                        [False, True, True, False, False],
-                        [False, True, True, False, False],
-                        [False, False, False, False, False],
-                        [False, False, False, False, False],
-                    ],
-                    [
-                        [False, False, False, False, False],
-                        [False, False, False, True, True],
-                        [False, False, False, True, True],
-                        [False, False, False, False, False],
-                        [False, False, False, False, False],
-                    ],
-                ]
-            ),
-            0.5,
-            np.array([False, True, True]),
-            DoesNotRaise(),
-        ),  # three masks with no category
-        (
-            np.array(
-                [
-                    [0, 0, 0, 0, 0.8, 0],
-                    [0, 0, 0, 0, 0.85, 1],
-                    [0, 0, 0, 0, 0.9, 2],
-                ]
-            ),
-            np.array(
-                [
-                    [
-                        [False, False, False, False, False],
-                        [False, True, True, False, False],
-                        [False, True, True, False, False],
-                        [False, False, False, False, False],
-                        [False, False, False, False, False],
-                    ],
-                    [
-                        [False, False, False, False, False],
-                        [False, True, True, False, False],
-                        [False, True, True, False, False],
-                        [False, True, True, False, False],
-                        [False, False, False, False, False],
-                    ],
-                    [
-                        [False, False, False, False, False],
-                        [False, True, True, False, False],
-                        [False, True, True, False, False],
-                        [False, False, False, False, False],
-                        [False, False, False, False, False],
-                    ],
-                ]
-            ),
-            0.5,
-            np.array([True, True, True]),
-            DoesNotRaise(),
-        ),  # three masks with different category
-    ],
-)
-def test_mask_non_max_suppression(
-    predictions: np.ndarray,
-    masks: np.ndarray,
-    iou_threshold: float,
-    expected_result: Optional[np.ndarray],
-    exception: Exception,
-) -> None:
-    with exception:
-        result = mask_non_max_suppression(
-            predictions=predictions, masks=masks, iou_threshold=iou_threshold
-        )
-        assert np.array_equal(result, expected_result)
-
-
 @pytest.mark.parametrize(
     "xyxy, resolution_wh, expected_result",
     [
@@ -1030,8 +720,8 @@ def test_calculate_masks_centroids(
         ),  # two data dicts with the same field name and different length arrays values
         (
             [{}, {"test_1": [1, 2, 3]}],
-            {"test_1": [1, 2, 3]},
-            DoesNotRaise(),
+            None,
+            pytest.raises(ValueError),
         ),  # two data dicts; one empty and one non-empty dict
         (
             [{"test_1": [], "test_2": []}, {"test_1": [1, 2, 3], "test_2": [1, 2, 3]}],
@@ -1265,3 +955,138 @@ def test_get_data_item(
                 assert (
                     result[key] == expected_result[key]
                 ), f"Mismatch in non-array data for key {key}"
+
+
+@pytest.mark.parametrize(
+    "mask, expected_result, exception",
+    [
+        (
+            np.array([[0, 0, 0, 0], [0, 1, 1, 0], [0, 1, 0, 0], [0, 1, 1, 0]]).astype(
+                bool
+            ),
+            False,
+            DoesNotRaise(),
+        ),  # foreground object in one continuous piece
+        (
+            np.array([[1, 0, 0, 0], [1, 0, 0, 0], [0, 0, 0, 0], [0, 1, 1, 0]]).astype(
+                bool
+            ),
+            False,
+            DoesNotRaise(),
+        ),  # foreground object in 2 seperate elements
+        (
+            np.array([[0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0]]).astype(
+                bool
+            ),
+            False,
+            DoesNotRaise(),
+        ),  # no foreground pixels in mask
+        (
+            np.array([[1, 1, 1, 1], [1, 1, 1, 1], [1, 1, 1, 1], [1, 1, 1, 1]]).astype(
+                bool
+            ),
+            False,
+            DoesNotRaise(),
+        ),  # only foreground pixels in mask
+        (
+            np.array([[1, 1, 1, 0], [1, 0, 1, 0], [1, 1, 1, 0], [0, 0, 0, 0]]).astype(
+                bool
+            ),
+            True,
+            DoesNotRaise(),
+        ),  # foreground object has 1 hole
+        (
+            np.array([[1, 1, 1, 0], [1, 0, 1, 1], [1, 1, 0, 1], [0, 1, 1, 1]]).astype(
+                bool
+            ),
+            True,
+            DoesNotRaise(),
+        ),  # foreground object has 2 holes
+    ],
+)
+def test_contains_holes(
+    mask: npt.NDArray[np.bool_], expected_result: bool, exception: Exception
+) -> None:
+    with exception:
+        result = contains_holes(mask)
+        assert result == expected_result
+
+
+@pytest.mark.parametrize(
+    "mask, connectivity, expected_result, exception",
+    [
+        (
+            np.array([[0, 0, 0, 0], [0, 1, 1, 0], [0, 1, 0, 0], [0, 1, 1, 0]]).astype(
+                bool
+            ),
+            4,
+            False,
+            DoesNotRaise(),
+        ),  # foreground object in one continuous piece
+        (
+            np.array([[1, 0, 0, 0], [1, 0, 0, 0], [0, 0, 0, 0], [0, 1, 1, 0]]).astype(
+                bool
+            ),
+            4,
+            True,
+            DoesNotRaise(),
+        ),  # foreground object in 2 seperate elements
+        (
+            np.array([[0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0], [0, 0, 0, 0]]).astype(
+                bool
+            ),
+            4,
+            False,
+            DoesNotRaise(),
+        ),  # no foreground pixels in mask
+        (
+            np.array([[1, 1, 1, 1], [1, 1, 1, 1], [1, 1, 1, 1], [1, 1, 1, 1]]).astype(
+                bool
+            ),
+            4,
+            False,
+            DoesNotRaise(),
+        ),  # only foreground pixels in mask
+        (
+            np.array([[1, 1, 1, 0], [1, 0, 1, 1], [1, 1, 0, 1], [0, 1, 1, 1]]).astype(
+                bool
+            ),
+            4,
+            False,
+            DoesNotRaise(),
+        ),  # foreground object has 2 holes, but is in single piece
+        (
+            np.array([[1, 1, 0, 0], [1, 1, 0, 1], [1, 0, 1, 1], [0, 0, 1, 1]]).astype(
+                bool
+            ),
+            4,
+            True,
+            DoesNotRaise(),
+        ),  # foreground object in 2 elements with respect to 4-way connectivity
+        (
+            np.array([[1, 1, 0, 0], [1, 1, 0, 1], [1, 0, 1, 1], [0, 0, 1, 1]]).astype(
+                bool
+            ),
+            8,
+            False,
+            DoesNotRaise(),
+        ),  # foreground object in single piece with respect to 8-way connectivity
+        (
+            np.array([[1, 1, 0, 0], [1, 1, 0, 1], [1, 0, 1, 1], [0, 0, 1, 1]]).astype(
+                bool
+            ),
+            5,
+            None,
+            pytest.raises(ValueError),
+        ),  # Incorrect connectivity parameter value, raises ValueError
+    ],
+)
+def test_contains_multiple_segments(
+    mask: npt.NDArray[np.bool_],
+    connectivity: int,
+    expected_result: bool,
+    exception: Exception,
+) -> None:
+    with exception:
+        result = contains_multiple_segments(mask=mask, connectivity=connectivity)
+        assert result == expected_result

test/utils/test_internal.py

@@ -0,0 +1,193 @@
+from contextlib import ExitStack as DoesNotRaise
+from dataclasses import dataclass, field
+from typing import Any, Set
+
+import numpy as np
+import pytest
+
+from supervision.detection.core import Detections
+from supervision.utils.internal import get_instance_variables
+
+
+class MockClass:
+    def __init__(self):
+        self.public = 0
+        self._protected = 1
+        self.__private = 2
+
+    def public_method(self):
+        pass
+
+    def _protected_method(self):
+        pass
+
+    def __private_method(self):
+        pass
+
+    @property
+    def public_property(self):
+        return 0
+
+    @property
+    def _protected_property(self):
+        return 1
+
+    @property
+    def __private_property(self):
+        return 2
+
+
+@dataclass
+class MockDataclass:
+    public: int = 0
+    _protected: int = 1
+    __private: int = 2
+
+    public_field: int = field(default=0)
+    _protected_field: int = field(default=1)
+    __private_field: int = field(default=2)
+
+    public_field_with_factory: dict = field(default_factory=dict)
+    _protected_field_with_factory: dict = field(default_factory=dict)
+    __private_field_with_factory: dict = field(default_factory=dict)
+
+    def public_method(self):
+        pass
+
+    def _protected_method(self):
+        pass
+
+    def __private_method(self):
+        pass
+
+    @property
+    def public_property(self):
+        return 0
+
+    @property
+    def _protected_property(self):
+        return 1
+
+    @property
+    def __private_property(self):
+        return 2
+
+
+@pytest.mark.parametrize(
+    "input_instance, include_properties, expected, exception",
+    [
+        (
+            MockClass,
+            False,
+            None,
+            pytest.raises(ValueError),
+        ),
+        (
+            MockClass(),
+            False,
+            {"public"},
+            DoesNotRaise(),
+        ),
+        (
+            MockClass(),
+            True,
+            {"public", "public_property"},
+            DoesNotRaise(),
+        ),
+        (
+            MockDataclass(),
+            False,
+            {"public", "public_field", "public_field_with_factory"},
+            DoesNotRaise(),
+        ),
+        (
+            MockDataclass(),
+            True,
+            {"public", "public_field", "public_field_with_factory", "public_property"},
+            DoesNotRaise(),
+        ),
+        (
+            Detections,
+            False,
+            None,
+            pytest.raises(ValueError),
+        ),
+        (
+            Detections,
+            True,
+            None,
+            pytest.raises(ValueError),
+        ),
+        (
+            Detections.empty(),
+            False,
+            {"xyxy", "class_id", "confidence", "mask", "tracker_id", "data"},
+            DoesNotRaise(),
+        ),
+        (
+            Detections.empty(),
+            True,
+            {
+                "xyxy",
+                "class_id",
+                "confidence",
+                "mask",
+                "tracker_id",
+                "data",
+                "area",
+                "box_area",
+            },
+            DoesNotRaise(),
+        ),
+        (
+            Detections(xyxy=np.array([[1, 2, 3, 4]])),
+            False,
+            {
+                "xyxy",
+                "class_id",
+                "confidence",
+                "mask",
+                "tracker_id",
+                "data",
+            },
+            DoesNotRaise(),
+        ),
+        (
+            Detections(
+                xyxy=np.array([[1, 2, 3, 4], [5, 6, 7, 8]]),
+                class_id=np.array([1, 2]),
+                confidence=np.array([0.1, 0.2]),
+                mask=np.array([[[1]], [[2]]]),
+                tracker_id=np.array([1, 2]),
+                data={"key_1": [1, 2], "key_2": [3, 4]},
+            ),
+            False,
+            {
+                "xyxy",
+                "class_id",
+                "confidence",
+                "mask",
+                "tracker_id",
+                "data",
+            },
+            DoesNotRaise(),
+        ),
+        (
+            Detections.empty(),
+            False,
+            {"xyxy", "class_id", "confidence", "mask", "tracker_id", "data"},
+            DoesNotRaise(),
+        ),
+    ],
+)
+def test_get_instance_variables(
+    input_instance: Any,
+    include_properties: bool,
+    expected: Set[str],
+    exception: Exception,
+) -> None:
+    with exception:
+        result = get_instance_variables(
+            input_instance, include_properties=include_properties
+        )
+        assert result == expected