supervision-0.11.0

🚀 Added

• Ability to load and save sv.DetectionDataset in COCO format using as_coco and from_coco methods. (#150)

>>> import supervision as sv

>>> ds = sv.DetectionDataset.from_coco(
...     images_directory_path='...',
...     annotations_path='...'
... )

>>> ds.as_coco(
...     images_directory_path='...',
...     annotations_path='...'
... )

• Ability to marge multiple sv.DetectionDataset together using merge method. (#158)

>>> import supervision as sv

>>> ds_1 = sv.DetectionDataset(...)
>>> len(ds_1)
100
>>> ds_1.classes
['dog', 'person']

>>> ds_2 = sv.DetectionDataset(...)
>>> len(ds_2)
200
>>> ds_2.classes
['cat']

>>> ds_merged = sv.DetectionDataset.merge([ds_1, ds_2])
>>> len(ds_merged)
300
>>> ds_merged.classes
['cat', 'dog', 'person']

• Additional start and end arguments to sv.get_video_frames_generator allowing to generate frames only for a selected part of the video. (#162)

🛠️ Fixed

• Incorrect loading of YOLO dataset class names from data.yaml. (#157)

🏆 Contributors

@SkalskiP @hardikdava