kwcoco.cli.coco_split module¶

class kwcoco.cli.coco_split.CocoSplitCLI(*args, **kwargs)[source]¶

Bases: DataConfig

Splits a COCO files into two or more parts base on some criteria.

Useful for generating quick and dirty train/test splits, but in general users should opt for using kwcoco subset instead to explicitly construct these splits based on domain knowledge.

Note

This may currently have a bug with split sizes. Help wanted.

Valid options: []

Parameters:

*args – positional arguments for this data config
**kwargs – keyword arguments for this data config

classmethod main(cmdline=True, **kw)[source]¶

CommandLine

xdoctest -m kwcoco.cli.coco_split CocoSplitCLI.main

Example

>>> from kwcoco.cli.coco_split import *  # NOQA
>>> import ubelt as ub
>>> dpath = ub.Path.appdir('kwcoco/tests/cli/split').ensuredir()
>>> kw = {'src': 'special:vidshapes8',
>>>       'dst1': dpath / 'train.json',
>>>       'dst2': dpath / 'test.json'}
>>> cmdline = False
>>> cls = CocoSplitCLI
>>> cls.main(cmdline, **kw)

default = {'balance_categories': <Value(True)>, 'compress': <Value('auto')>, 'dst1': <Value('split1.kwcoco.json')>, 'dst2': <Value('split2.kwcoco.json')>, 'factor': <Value(3)>, 'num_write': <Value(1)>, 'rng': <Value(None)>, 'splitter': <Value('auto')>, 'src': <Value(None)>, 'test_fpath': <Value(None)>, 'train_fpath': <Value(None)>, 'vali_fpath': <Value(None)>}¶