Language

unilab.tasks.motion_tracking.common.motion_loader.MotionLoader¶

class unilab.tasks.motion_tracking.common.motion_loader.MotionLoader[source]¶

Bases: object

Loads and provides access to motion data from NPZ files.

Parameters:

motion_file (str | Sequence[str])
body_indices (Sequence[int] | None)

Methods

`__init__`(motion_file[, body_indices])	Initialize motion loader.
`clip_end_frames_torch`(device)	Return immutable clip end frames as a device-resident Torch table.
`clip_offsets_torch`(device)	Return immutable clip start frames as a device-resident Torch table.
`get_clip_indices`(frame_idx)	Map global frame indices to clip indices.
`motion_features_torch`(device)	Return the immutable motion feature table on the requested device.

__init__(motion_file, body_indices=None)[source]¶

Initialize motion loader.

Parameters:

motion_file (str | Sequence[str]) – Path to one NPZ file, or a sequence of NPZ files
body_indices (Sequence[int] | None) – Optional indices into the NPZ body axis. The exported motion files currently keep MuJoCo body-id layout, so these indices are expected to follow that convention.

get_clip_indices(frame_idx)[source]¶

Map global frame indices to clip indices.

Parameters:: frame_idx (ndarray)
Return type:: ndarray

motion_features_torch(device)[source]¶

Return the immutable motion feature table on the requested device.

NPZ parsing and validation remain inside this cold-path leaf. Runtime consumers receive one contiguous float32 table with columns packed in the documented motion-feature order rather than mutable NumPy arrays.

Parameters:: device (str | device)
Return type:: Tensor

clip_offsets_torch(device)[source]¶

Return immutable clip start frames as a device-resident Torch table.

Parameters:: device (str | device)
Return type:: Tensor

clip_end_frames_torch(device)[source]¶

Return immutable clip end frames as a device-resident Torch table.

Parameters:: device (str | device)
Return type:: Tensor