pablovela5620/assembly101-720p

Assembly101 720p AV1 SimpleCV Variant

This is an unofficial 720p AV1 SimpleCV-compatible variant of the Assembly101 dataset. It is not the official full-resolution upstream release. The official upstream Hugging Face dataset is cvml-nus/assembly101.

Contents

• Local video sequences: 354
• Local MP4 files: 4248
• SimpleCV-ingestable sequences: 337
• Video-only sequences: 17
• Sequences with existing canonical RRD baseline: 334

The repository preserves the current SimpleCV-compatible folder layout so callers can use:

from pathlib import Path
from simplecv.data.exoego.assembly101 import Assembly101Config, Assembly101Sequence

config = Assembly101Config(
    root_directory=Path("/path/to/downloaded/assembly101-720p"),
    sequence_name="<sequence_name>",
)
sequence = Assembly101Sequence(config)

Important Notes

• All 354 local 720p AV1 video sequence folders are included.
• 337 sequences include the pose and camera metadata required by current SimpleCV ingestion.
• 17 sequences are included as video-only assets and are marked in manifests/sequences.csv.
• The extracted payload is intentionally reduced to SimpleCV-required annotations and calibration.
• For archival completeness, the official AssemblyPoses.zip (72 GB, byte-identical copy of cvml-nus/assembly101's archive, added 2026-07-03) is included at the repo root. It contains all ten pose/calibration components (landmarks2D, landmarks3D, hand_bboxes, hand_confidences, camera_extrinsics_{ego,fixed}, camera_position_{ego,fixed}, timestamp, xf_transf) — a superset of the extracted subset below.
• The complete AssemblyHands v1-1 annotation JSONs (train + val: ego/exo calib, data, joint_3d) are included under assembly-hands/annotations/.
• Full upstream 1080p videos and full toolkit source are still not included (see cvml-nus/assembly101 and facebookresearch/assemblyhands-toolkit).
• This dataset is licensed under CC BY-NC 4.0 and is intended for noncommercial research use.

Layout

AssemblyPoses.zip                                    # official full pose archive (72 GB)
videos/av1-720-new/
assembly101_camera_and_hand_poses/landmarks3D/
assembly101_camera_and_hand_poses/camera_extrinsics_ego/
assembly101_camera_and_hand_poses/camera_extrinsics_fixed/
assembly-hands/annotations/train/                    # calib, data, joint_3d (v1-1)
assembly-hands/annotations/val/                      # calib (ego+exo), data, joint_3d (v1-1)
assemblyhands-toolkit/calib/nimble_json_calib/
manifests/

Manifests

• manifests/sequences.csv: one row per local 720p video sequence.
• manifests/upload_payload.txt: include patterns used for the large upload.
• manifests/smoke_test_sequences_10_seed_20260517.txt: fixed-seed sample for the final SimpleCV/Rerun smoke test.

Citation

Please cite Assembly101:

@article{sener2022assembly101,
    title = {Assembly101: A Large-Scale Multi-View Video Dataset for Understanding Procedural Activities},
    author = {F. Sener and D. Chatterjee and D. Shelepov and K. He and D. Singhania and R. Wang and A. Yao},
    journal = {CVPR 2022},
}

Please cite AssemblyHands when using the hand-pose annotations and calibration files:

@inproceedings{ohkawa2023assemblyhands,
    title = {{AssemblyHands:} Towards Egocentric Activity Understanding via 3D Hand Pose Estimation},
    author = {Takehiko Ohkawa and Kun He and Fadime Sener and Tomas Hodan and Luan Tran and Cem Keskin},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    pages = {12999-13008},
    year = {2023},
}

License

Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0). See LICENSE.