support for yolo dataset loading and saving

[SkalskiP]

Description

YOLO Dataset loader is designed to be as flexible as possible to allow data consistency for both mixed and non-mixed datasets. You can test API in this example notebook.

• box dataset

• mask dataset

• non-mixed datasets

API

• To load a consistent YOLO dataset. Only boxes or boxes and polygons.

dataset = sv.Dataset.from_yolo(
    images_directory_path=f"{dataset.location}/train/images",
    annotations_directory_path=f"{dataset.location}/train/labels",
    data_yaml_path=f"{dataset.location}/data.yaml",
    force_masks=False
)

• Force consistent mask loading from mixed-label YOLO dataset.

dataset = sv.Dataset.from_yolo(
    images_directory_path=f"{dataset.location}/train/images",
    annotations_directory_path=f"{dataset.location}/train/labels",
    data_yaml_path=f"{dataset.location}/data.yaml",
    force_masks=True
)

[Jacobsolawetz]

🔥 splits are separate sv Datasets?

[SkalskiP]

Yeah. You load every split as a separate sv.Dataset but that is pretty much the standard approach. Torch Datasets and Dataloaders work the same way. But I think we will build some utils to split and merge datasets to.