fastdev.datasets.dexycb

Module Contents

class fastdev.datasets.dexycb.DexYCBDatasetConfig

Configuration for DexYCBDataset.

data_root: str

hand_side: Literal['left', 'right', 'both'] = 'left'

setup: Literal['s0', 's1', 's2', 's3'] = 's0'

• [s0] Seen subjects, camera views, grasped objects, unseen frames

• [s1] Unseen subjects

• [s2] Unseen camera views

• [s3] Unseen grasped objects

Reference

serials: List[str] | None = None

return_color: bool = False

return_depth: bool = False

return_object_mesh: bool = False

return_as_sequence: bool = False

return_no_hand_frames: bool = True

download_simplified_data_if_not_exist: bool = False

force_rebuild_cache: bool = False

__post_init__() → None

Return type:

None

class fastdev.datasets.dexycb.DexYCBDataset(config: DexYCBDatasetConfig, split: Literal['train', 'val', 'test', 'all'] = 'train')

Bases: torch.utils.data.Dataset

DexYCBDataset.

Parameters:

• config (DexYCBDatasetConfig) – configuration for the dataset.

• split (Literal['train', 'val', 'test', 'all'])

Examples

>>> from fastdev.datasets.dexycb import DexYCBDataset, DexYCBDatasetConfig
>>> dataset = DexYCBDataset(DexYCBDatasetConfig())

ycb_classes

config

split = 'train'

__len__()

__getitem__(idx)

property h

property w

property subjects

property serials

property sequences

select_indices(seq: str, serial: str | None = None)

Parameters:

• seq (str)

• serial (Optional[str])

build_collate_fn()

static download_data(data_root: str)

Parameters:

data_root (str)