keypoints_to_detections and Keypoint tracking

[LinasKo]

Description

The simplest way to implement keypoint tracking is to reuse the functionality in detections. In this PR I add:

• keypoints_to_detections function for converstion between the two types
• Docs, guides and examples how keypoint tracking can be done

Minors changes:

• Added KeyPoints.is_empty()
• typo & format fixes

🟢 This approach does not lock us in whatsoever - we can easily introduce a better tracking system later, without deprecations.

track-keypoints.mp4

Type of change

Please delete options that are not relevant.

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• This change requires a documentation update

How has this change been tested, please provide a testcase or example of how you tested the change?

1. Tested in Colab: https://colab.research.google.com/drive/1Rh7CRUYZK7xLioOVYAQPQTENSrOqTmfe?usp=sharing
2. Generated example video
3. Looked through the docs repeatedly

Generate example video for docs

I used a smaller, 1280 x 720 video.
https://www.pexels.com/video/ski-montagne-skier-piste-de-ski-4274798/

import numpy as np
import supervision as sv
from ultralytics import YOLO

model = YOLO("yolov8m-pose.pt")
tracker = sv.ByteTrack()
smoother = sv.DetectionsSmoother()
box_annotator = sv.BoundingBoxAnnotator()
label_annotator = sv.LabelAnnotator()
trace_annotator = sv.TraceAnnotator()

def callback(frame: np.ndarray, _: int) -> np.ndarray:
    results = model(frame)[0]
    keypoints = sv.KeyPoints.from_ultralytics(results)
    detections = sv.keypoints_to_detections(keypoints)
    detections = tracker.update_with_detections(detections)
    detections = smoother.update_with_detections(detections)

    labels = [
        f"#{tracker_id} {results.names[class_id]}"
        for class_id, tracker_id
        in zip(detections.class_id, detections.tracker_id)
    ]

    annotated_frame = box_annotator.annotate(
        frame.copy(), detections=detections)
    annotated_frame = label_annotator.annotate(
        annotated_frame, detections=detections, labels=labels)
    return trace_annotator.annotate(
        annotated_frame, detections=detections)

sv.process_video(
    source_path="skiing2.mp4",
    target_path="result.mp4",
    callback=callback
)

Any specific deployment considerations

Docs

• Docs updated? What were the changes:

• New section in "Track Objects on Video"
• New subsection "Datatypes" in supervision global utils.

[LinasKo]

Future improvement: Show keypoints in end video. This likely needs new sections - detecting and drawing keypoints, then converting and also showing boxes, and then doing tracking.