concepts.benchmark.vision_language.babel_qa.dataset.BabelQADatasetUnwrapped

class BabelQADatasetUnwrapped

Bases: FilterableDatasetUnwrapped

BabelQA dataset.

Methods

get_metainfo(index)

__add__(other)

__getitem__(index)

__init__(data_dir, data_split_file, split, data_source, no_gt_segments, filter_supervision, max_frames=150)

Initialize BabelQA dataset.

Parameters:

• data_dir (str) – path to data directory.

• data_split_file (str) – path to the json file that contains the question ids for each split.

• split (str) – split to use.

• data_source (str) – data source to use. Either teach_synth or ???

• no_gt_segments (bool) – whether to use ground truth segments or not.

• filter_supervision –

  ???

• max_frames (int) – maximum number of frames to use per segment. (is this correct???)

__iter__()

__len__()

__new__(**kwargs)

get_metainfo(index)