Dataset Viewer
Duplicate
The dataset viewer is not available for this split.
Job manager crashed while running this job (missing heartbeats).
Error code:   JobManagerCrashedError

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

bible-nlp/sign-bibles

This dataset contains sign language videos from the Digital Bible Library (DBL), processed for machine learning applications. The dataset is licensed under the Creative Commons Attribution-ShareAlike 4.0 International License (CC BY-SA 4.0).

Dataset Structure

Each sample contains:

  • ["mp4"] the original video
  • ["json"] Metadata, including bible reference, copyright information, language information.
  • ["transcripts.json"]: Bible verses from the eBible corpus, as well as vref indices. For some videos, more fine-grained annotations are available with start/end frames for certain sections
  • ["pose-mediapipe.pose"]: Mediapipe holistic keypoints, created with Pose Format
  • ["pose-dwpose.npz"]: DWPose outputs, saved as compressed npz files.

Usage (Tested on Linux)

Requires installing torchcodec, torchvision

import argparse
import io
import json
from pathlib import Path

import pandas as pd
from datasets import Video, get_dataset_config_names, load_dataset
from pose_format import Pose
from torchvision.io import write_png
from tqdm import tqdm


def iterate_over_dataset(language_subset: str, sample_count: int) -> pd.DataFrame:
    """Extract valid queries from the Sign Bibles dataset."""
    # https://huggingface.co/docs/datasets/en/video_load#webdataset
    # The dataset is massive! Best to stream it
    ds = load_dataset("bible-nlp/sign-bibles", language_subset, streaming=True, split="train")

    print(f"Loading dataset subset '{language_subset}: {ds}'")
    ds = ds.cast_column("mp4", Video())  # for convenience, otherwise we just get Value('binary'), aka just bytes

    for sample in tqdm(ds.take(sample_count), desc="Parsing samples"):
        sample_key = sample["__key__"]
        # load metadata
        metadata = sample["json"]
        print(json.dumps(metadata, indent=2))
        total_frames = metadata["total_frames"]  # total frame count of the video

        # load Pose format. Normally it expects a file buffer, so..
        pose = Pose.read(io.BytesIO(sample["pose-mediapipe.pose"]))
        print("Pose Format (https://github.com/sign-language-processing/pose)")
        print(pose)

        # DWPose (not present for every sample)
        if "pose-dwpose.npz" in sample:
            dwpose = sample["pose-dwpose.npz"]
            print("DWPose Format")
            print(type(dwpose))  # dictionary with "frames" as the key, sometimes also "confidences"

        # load video
        video = sample["mp4"]
        print(type(video))  # bytes
        first_frame = video.get_frame_at(0)  # torchcodec Frame type
        print(type(first_frame))
        print(first_frame.data.shape)

        # load transcripts and get start/end frames for a segment
        if "transcripts.json" in sample:
            transcripts = sample["transcripts.json"]
            print(f"Transcripts for {sample_key}")
            for transcript in transcripts:
                start_frame_index = transcript["start_frame"]

                end_frame_index = transcript["end_frame"]
                print(transcript["text"][:100], start_frame_index, end_frame_index)

                # torchcodec._frame.Frame's .data has the data
                start_frame = video.get_frame_at(start_frame_index).data
                end_frame = video.get_frame_at(end_frame_index).data

                start_out_path = f"{sample_key}_{start_frame_index}.png"
                print(f"Writing to {Path(start_out_path).resolve()}")
                # write_png(start_frame, start_out_path)

                end_out_path = f"{sample_key}_{end_frame_index}.png"
                print(f"Writing to {Path(end_out_path).resolve()}")
                # write_png(end_frame, end_out_path)


def main(
    language_subset: str,
    sample_count: int,
):
    iterate_over_dataset(language_subset=language_subset, sample_count=sample_count)


if __name__ == "__main__":
    parser = argparse.ArgumentParser(description="Extract Sign Bibles queries from Huggingface dataset.")
    parser.add_argument(
        "--language-subset",
        type=str,
        default="ase",
        help="Language subset to download (default: ase)",
    )
    parser.add_argument("--sample-count", type=int, default=1)

    args = parser.parse_args()

    if args.language_subset is not None:
        configs = [args.language_subset]
    else:
        configs = get_dataset_config_names("bible-nlp/sign-bibles")
        print(f"Available configs: {configs}")

    for language_subset in configs:
        main(language_subset=language_subset, sample_count=args.sample_count)

Known Issues

  • Videos larger than 2 GB are not in this repository. Loading them with datasets fails with a PyArrow ArrowCapacityError (huggingface/datasets#7691), so they were removed from the uploaded shards. A copy of the full WebDataset, including these videos, is available on Box. See BibleNLP/sign-bibles-dataset#22.
  • The Dataset Viewer is not available. The shards are too large for the Hub's viewer. Stream the dataset instead, as shown under Usage.
  • Some videos have non-square pixels. Frames extracted without applying the sample aspect ratio are horizontally squeezed; keypoints computed from such frames may be affected too. See BibleNLP/sign-bibles-dataset#45.

Gloss Annotations

Timestamped gloss annotations by a Deaf annotator for 10 videos from the American Sign Language subset (679 annotations) are in the osb_glosses folder of this repository (osb_glosses_v1_4.parquet, CC BY-SA 4.0).

Citation

If you use this dataset, please cite the preprint (doi:10.21203/rs.3.rs-7918028/v1):

@misc{leong2026osb,
  title        = {Open Sign Bibles ({OSB}): An Open-Access Multilingual Sign Language Bibles Dataset},
  author       = {Leong, Colin and Nemecek, Joshua and Raju, Kavitha and Mathew, Joel and Asari, Vijayan and Leong, Elisabeth},
  year         = {2026},
  howpublished = {Research Square preprint},
  doi          = {10.21203/rs.3.rs-7918028/v1}
}

License and Attribution

This dataset is derived from the Digital Bible Library (DBL) sign language content. Each sample includes copyright and license information in its metadata.

The dataset is provided under a CC BY-SA 4.0 license, which allows for:

  • Sharing: Copy and redistribute the material in any medium or format
  • Adaptation: Remix, transform, and build upon the material

Under the following terms:

  • Attribution: You must give appropriate credit, provide a link to the license, and indicate if changes were made
  • ShareAlike: If you remix, transform, or build upon the material, you must distribute your contributions under the same license as the original

Please refer to the individual sample metadata for specific attribution requirements.

Downloads last month
3,720
Size
: 1653 shards, approximately 8105 samples
Languages
: ase, bqn, esl, eso, eth, gse, ins, lsb, mis, mzy, nsi, nsp, rsl, sqs, tza, ugn, xki
Projects
: Chronological Bible Translation in American Sign Language (119 Introductions and Passages expanded with More Information), Chronological Bible Translation in American Sign Language (119 Introductions and Passages), Chronological Bible Translation in Andhra Pradesh Sign Language, Chronological Bible Translation in Bulgarian Sign Language, Chronological Bible Translation in Burundian Sign Language, Chronological Bible Translation in Egyptian Sign Language, Chronological Bible Translation in Ethiopian Sign Language, Chronological Bible Translation in Ghanaian Sign Language, Chronological Bible Translation in Indian (Delhi) Sign Language, Chronological Bible Translation in Kenyan Sign Language, Chronological Bible Translation in Kerala Sign Language, Chronological Bible Translation in Mozambican Sign Language, Chronological Bible Translation in Nepali Sign Language, Chronological Bible Translation in Nigerian Sign Language, Chronological Bible Translation in Russian Sign Language, Chronological Bible Translation in South Sudanese Sign Language, Chronological Bible Translation in Sri Lankan Sign Language, Chronological Bible Translation in Tanzanian Sign Language, Chronological Bible Translation in Ugandan Sign Language, Chronological Bible Translation in West Bengal Sign Language, The Bible in Estonian Sign Language
License
CC BY-SA 4.0
Number of rows:
150
Total file size:
2.39 TB