Add handball court support

examples/handball/.gitignore

@@ -0,0 +1,4 @@
+data/
+notebooks/*.pt
+notebooks/*/
+notebooks/runs/

examples/handball/README.md

@@ -0,0 +1,111 @@
+# Handball AI
+
+## install
+
+Install from source in a [Python>=3.8](https://www.python.org/) environment.
+
+```bash
+pip install git+https://github.com/roboflow/sports.git
+cd examples/handball
+pip install -r requirements.txt
+./setup.sh
+```
+
+If a notebook fails with `RuntimeError: Numpy is not available`, restart the
+runtime after the install cell finishes, then run the notebook from the top.
+This is caused by PyTorch importing before the notebook pins a NumPy version.
+
+## datasets
+
+| use case                         | dataset                                                                                                                                             | train model                                                                                                                                            |
+|:---------------------------------|:----------------------------------------------------------------------------------------------------------------------------------------------------|:-------------------------------------------------------------------------------------------------------------------------------------------------------|
+| handball player detection        | [Handball_Player_Detection](https://universe.roboflow.com/project-k1mun/handball_player_detection)                                                  | [train_player_detector.ipynb](notebooks/train_player_detector.ipynb)                                                                                   |
+| handball ball detection          | [handball-ballon](https://universe.roboflow.com/handballdetection/handball-ballon)                                                                  | [train_ball_detector.ipynb](notebooks/train_ball_detector.ipynb)                                                                                       |
+| handball ball detection          | [ball-only-detection](https://universe.roboflow.com/handball-detection/ball-only-detection)                                                        | [train_ball_detector.ipynb](notebooks/train_ball_detector.ipynb)                                                                                       |
+| handball court keypoint detection | [Handball Keypoint Detection](https://universe.roboflow.com/handball-wckh8/handball-keypoint-detection-ndvsl)                                      | [train_court_keypoint_detector.ipynb](notebooks/train_court_keypoint_detector.ipynb)                                                                   |
+
+After training, place the exported model weights in:
+
+```text
+examples/handball/data/handball-player-detection.pt
+examples/handball/data/handball-ball-detection.pt
+examples/handball/data/handball-court-keypoint-detection.pt
+```
+
+## modes
+
+- `COURT_RENDERING` - Renders an IHF-standard handball court with the 4 m
+  throw-off area, 6 m goal-area lines, 9 m free-throw lines, 7 m lines,
+  4 m goalkeeper restraining lines, and substitution marks.
+
+  ```bash
+  python main.py --target_dir data --mode COURT_RENDERING
+  ```
+
+- `POINT_RENDERING` - Draws sample player and ball positions on the handball
+  court.
+
+  ```bash
+  python main.py --target_dir data --mode POINT_RENDERING
+  ```
+
+- `PATH_RENDERING` - Draws sample movement paths on the handball court.
+
+  ```bash
+  python main.py --target_dir data --mode PATH_RENDERING
+  ```
+
+- `COURT_DETECTION` - Detects handball court keypoints in a video.
+
+  ```bash
+  python main.py --source_video_path data/input.mp4 \
+  --target_video_path data/court-detection.mp4 \
+  --device cpu --mode COURT_DETECTION
+  ```
+
+- `PLAYER_DETECTION` - Detects players, referees, and goalkeepers in a video.
+
+  ```bash
+  python main.py --source_video_path data/input.mp4 \
+  --target_video_path data/player-detection.mp4 \
+  --device cpu --mode PLAYER_DETECTION
+  ```
+
+- `BALL_DETECTION` - Detects the handball in a video.
+
+  ```bash
+  python main.py --source_video_path data/input.mp4 \
+  --target_video_path data/ball-detection.mp4 \
+  --device cpu --mode BALL_DETECTION
+  ```
+
+- `RADAR` - Combines court keypoints and player detections to draw detected
+  player locations on a rendered handball court.
+
+  ```bash
+  python main.py --source_video_path data/input.mp4 \
+  --target_video_path data/radar.mp4 \
+  --device cpu --mode RADAR
+  ```
+
+- `ALL_RENDERINGS` - Runs all static rendering modes.
+
+  ```bash
+  python main.py --target_dir data --mode ALL_RENDERINGS
+  ```
+
+Static rendering outputs are written to `examples/handball/data`.
+
+## notebooks
+
+| use case                         | notebook                                                        |
+|:---------------------------------|:----------------------------------------------------------------|
+| handball player detection        | [train_player_detector.ipynb](notebooks/train_player_detector.ipynb) |
+| handball ball detection          | [train_ball_detector.ipynb](notebooks/train_ball_detector.ipynb) |
+| handball court keypoint detection | [train_court_keypoint_detector.ipynb](notebooks/train_court_keypoint_detector.ipynb) |
+| handball court render            | [render_court.ipynb](notebooks/render_court.ipynb)              |
+
+## license
+
+This example uses the `sports` package and Supervision-based rendering utilities,
+which are distributed under the MIT license.