TransformedFrameDataBatch

class standard_e2e.data_structures.TransformedFrameDataBatch(frames, *, device=None)

Bases: object

Data structure to hold a batch of frame data (PyTorch-friendly).

Parameters:

• frames (list[TransformedFrameData])

• device (device | None)

aux_data: Dict[str, Any] | None = None

cuda(device=None)

Move batched tensors and nested modality payloads to a CUDA device.

Parameters:

device (int | None)

dataset_name: list[str]

filename: list[str]

frame_id: list[int]

get_modality_data(modality)

Get data for a specific modality.

Return type:

Any

Parameters:

modality (Modality)

segment_id: list[str]

split: list[str]

timestamp: Tensor

timestamp_diff: Tensor | None = None

to(device)

Move all tensor-like fields to device (non-blocking where possible).

Parameters:

device (device)