[KeypointsDataset] - create a new dataset format to load keypoints

[SkalskiP]

Description

Supervision can currently load object detection, instance segmentation, and classification datasets. Let's add the ability to handle keypoint datasets. The sv.KeypointDataset should offer the same functionalities as sv.DetectionDataset and sv.ClassificationDataset, including merge, split, __getitem__, __iter__, from_yolo, and as_yolo.

Additional

• Note: Please share a Google Colab with minimal code to test the new feature. We know it's additional work, but it will speed up the review process. The reviewer must test each change. Setting up a local environment to do this is time-consuming. Please ensure that Google Colab can be accessed without any issues (make it public). Thank you! 🙏🏻

[shaddu]

Hi @SkalskiP 👋

I was analyzing the issue and understood that KeypointsDataset will be similar to DetectionDataset, with the difference being the type of annotations.

classes: List[str],
images: Union[List[str], Dict[str, np.ndarray]],
annotations: Dict[str, KeyPoints]

If this is a correct understanding then I will like to work on this issue.

[LinasKo]

Let's only implement the new, lazy-loaded, non-deprecated API, with images: List[str]. This will save work in the future.

[LinasKo]

@shaddu, I also made this template, which should help with starting your colab: Colab Template.
Feel free to use as much or as little of it as you wish 😉

[shaddu]

Hi @LinasKo 👋

I also noticed the deprecated warning in the Datasets classes, and the clarification around images: List[str] is helpful.

Thank you for sharing the Colab template 😎 . It will be helpful, and I will suggest adding some similar Colab templates to the contribution guide.