Hugging Face's logo

Datasets:
YaojieShen
/
hhtools_omomo

Tasks:
Robotics
Languages:
English
Chinese
Size:
10K<n<100K
Tags:
motion-capture
human-object-interaction
omomo
humanoid
retargeting
mesh
License:
Dataset card Files
xet
 Community
Dataset Viewer
The dataset viewer is not available for this subset.
Cannot get the split names for the config 'default' of the dataset.
Exception:    SplitsNotFoundError
Message:      The split names could not be parsed from the dataset config.
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
                  for split_generator in builder._split_generators(
                                         ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/webdataset/webdataset.py", line 83, in _split_generators
                  raise ValueError(
                  ...<2 lines>...
                  )
              ValueError: The TAR archives of the dataset should be in WebDataset format, but the files in the archive don't share the same prefix or the same types.
              
              The above exception was the direct cause of the following exception:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 66, in compute_split_names_from_streaming_response
                  for split in get_dataset_split_names(
                               ~~~~~~~~~~~~~~~~~~~~~~~^
                      path=dataset,
                      ^^^^^^^^^^^^^
                      config_name=config,
                      ^^^^^^^^^^^^^^^^^^^
                      token=hf_token,
                      ^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
                  info = get_dataset_config_info(
                      path,
                  ...<6 lines>...
                      **config_kwargs,
                  )
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
                  raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
              datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.

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.

hhtools OMOMO — human-object interaction motion clips

中文说明

OMOMO-style human-object interaction motion data prepared for human-humanoid-tools (hhtools). The release contains clip-level windows and full-sequence folders, each paired with the corresponding object mesh. It is suitable for human-object interaction analysis, humanoid motion retargeting, and interaction-aware motion imitation workflows.

Clip folders 16,927
Full-sequence folders 5,882
Archives OMOMO_hhtools_all_clips.tar.gz, OMOMO_hhtools_all_full.tar.gz
Archive size ~2.7 GB total
Layout OMOMO-style per-sequence folders
Object assets Wavefront OBJ meshes, hard-linked when repeated
Source Converted from OMOMO human-object interaction data

Directory layout

This dataset is distributed as two compressed archives:

OMOMO_hhtools_all_clips.tar.gz
OMOMO_hhtools_all_full.tar.gz

After extraction, each sequence folder is self-contained:

<sequence_name>/
  <sequence_name>.pkl                 # processed motion / interaction data
  <object_name>_cleaned_simplified.obj # static object mesh

Examples:

OMOMO_hhtools_all_clips/
  sub10_clothesstand_000__train_w00000_f000000-000119/
    sub10_clothesstand_000__train_w00000_f000000-000119.pkl
    clothesstand_cleaned_simplified.obj

OMOMO_hhtools_all_full/
  sub13_vacuum_077__train_full/
    sub13_vacuum_077__train_full.pkl
    vacuum_cleaned_simplified.obj

Why archives instead of expanded folders?

The original processed folders use hard links for repeated object mesh files. Multiple sequence folders may point to the same underlying mesh on disk. This keeps local disk usage compact.

If the dataset is uploaded or downloaded as fully expanded files, those hard links may be materialized as duplicated files, making the apparent size much larger. The .tar.gz archives preserve hard links, so extracting with standard tar restores the compact layout.

Download 

# Hugging Face CLI
hf download YaojieShen/hhtools_omomo OMOMO_hhtools_all_clips.tar.gz --repo-type dataset
hf download YaojieShen/hhtools_omomo OMOMO_hhtools_all_full.tar.gz --repo-type dataset

Or download into a target directory:

mkdir -p ./hhtools_omomo
hf download YaojieShen/hhtools_omomo OMOMO_hhtools_all_clips.tar.gz --repo-type dataset --local-dir ./hhtools_omomo
hf download YaojieShen/hhtools_omomo OMOMO_hhtools_all_full.tar.gz --repo-type dataset --local-dir ./hhtools_omomo

Extract

Use regular tar extraction:

cd ./hhtools_omomo
tar -xzf OMOMO_hhtools_all_clips.tar.gz
tar -xzf OMOMO_hhtools_all_full.tar.gz

Do not extract with options or tools that dereference hard links if you want to preserve compact disk usage.

You can check the difference between real disk usage and apparent expanded size with:

du -sh OMOMO_hhtools_all_clips OMOMO_hhtools_all_full
du -shl --apparent-size OMOMO_hhtools_all_clips OMOMO_hhtools_all_full

Usage with hhtools

hhtools provides tools for humanoid motion retargeting and dataset inspection.

git clone https://github.com/jaggerShen/human-humanoid-tools.git
cd human-humanoid-tools
uv sync --extra all
uv run hhtools web

In the Web UI, point the motion library at the extracted OMOMO_hhtools_all_clips or OMOMO_hhtools_all_full folder, depending on whether you want windowed clips or complete sequences.

Motion / object categories

Folder names encode subject id, object category, sequence id, split, and clip window when applicable. Example object categories include:

  • Furniture / support objects — chair, table, largetable, clothesstand
  • Containers / carried objects — box, largebox, suitcase
  • Household tools — vacuum, mop, and related object interaction sequences
  • Windowed clips — names with wXXXXX_fSTART-END
  • Full sequences — names ending with train_full

Provenance & license

  • Upstream data: OMOMO human-object interaction motion data.
  • This release: OMOMO-derived data reorganized into hhtools-compatible per-sequence folders, with simplified object meshes packaged as hard-link-preserving archives.
  • License: CC-BY-NC-4.0 for this dataset release.

Please cite the original OMOMO work when using the motions in research, and cite or link hhtools if you use this packaging or the retargeting / analysis toolchain.

@software{human_humanoid_tools2026,
  title   = {human-humanoid-tools (hhtools): humanoid motion retargeting and dataset analysis},
  author  = {jaggerShen and hhtools contributors},
  year    = {2026},
  url     = {https://github.com/jaggerShen/human-humanoid-tools}
}

中文说明

本仓库提供面向 human-humanoid-tools (hhtools) 的 OMOMO 人-物交互动作数据。数据分为窗口化片段和完整序列两部分，每个序列文件夹包含处理后的动作 / 交互数据以及对应的物体 OBJ 网格。

数据概览

项目 数值
clip 片段数 16,927
full 完整序列数 5,882
文件包 OMOMO_hhtools_all_clips.tar.gz、OMOMO_hhtools_all_full.tar.gz
压缩包大小 总计约 2.7 GB
目录格式 OMOMO-style per-sequence folders
物体资产 OBJ 网格，重复物体使用 hard link 节省空间

目录结构 

<sequence_name>/
  <sequence_name>.pkl                 # 处理后的动作 / 交互数据
  <object_name>_cleaned_simplified.obj # 静态物体网格

下载 

mkdir -p ./hhtools_omomo
hf download YaojieShen/hhtools_omomo OMOMO_hhtools_all_clips.tar.gz --repo-type dataset --local-dir ./hhtools_omomo
hf download YaojieShen/hhtools_omomo OMOMO_hhtools_all_full.tar.gz --repo-type dataset --local-dir ./hhtools_omomo

解压 

cd ./hhtools_omomo
tar -xzf OMOMO_hhtools_all_clips.tar.gz
tar -xzf OMOMO_hhtools_all_full.tar.gz

请使用普通 tar 解压，不要使用会展开或复制 hard link 的选项。这样可以恢复本地的紧凑布局，避免重复物体网格占用大量磁盘空间。

在 hhtools 中使用

下载并解压后，启动 hhtools Web UI，将动作库路径指向 OMOMO_hhtools_all_clips 或 OMOMO_hhtools_all_full 文件夹即可。

来源与引用

数据源自 OMOMO 人-物交互数据，经整理为 hhtools 可直接读取的 per-sequence 文件夹格式，并以保留 hard link 的 .tar.gz 形式发布。使用时请引用原始 OMOMO 工作；如使用本整理格式或 hhtools 工具链，也请注明 hhtools。

Contact

Issues about loading / retargeting: human-humanoid-tools Issues

Dataset hosting: YaojieShen/hhtools_omomo
Downloads last month
45
Total file size:
2.82 GB