README.md

---
pretty_name: MotionMillion
size_categories:
- n<1T
task_categories:
- other
language:
- en
tags:
- Large Human Motion
- Humanoid
- Humanoid Locomotion
extra_gated_prompt: >-
  ### MotionMillion COMMUNITY LICENSE AGREEMENT

  MotionMillion Release Date: July 30, 2025 All the data and code
  within this repo are under [CC BY-NC-SA
  4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/).
extra_gated_fields:
  First Name: text
  Last Name: text
  Email: text
  Country: country
  Affiliation: text
  Phone: text
  Job title:
    type: select
    options:
    - Student
    - Research Graduate
    - AI researcher
    - AI developer/engineer
    - Reporter
    - Other
  Research interest: text
  geo: ip_location
  By clicking Submit below I accept the terms of the license and acknowledge that the information I provide will be collected stored processed and shared in accordance with the InternData Privacy Policy: checkbox
extra_gated_description: >-
  The information you provide will be collected, stored, processed and shared in
  accordance with the InternData Privacy Policy.
extra_gated_button_content: Submit
---

# 🔑 Key Features
- **Over 2000 hours of high-quality human motion** captured from web-scale human video data, covering:
  - Martial Arts (23.7%)
  - Fitness (26.4%)
  - Performance (17.5%)
  - Dance (14.9%)
  - Non-Human (2.9%)
  - Sports (2.4%)
- **Over 20 detailed annotations per motion**, including:
  - Age
  - Body Characteristics
  - Movement Styles
  - Emotions
  - Environments

<img src="assets/motionmillion_teaser.png" alt="Image Alt Text" width="70%" style="display: block; margin-left: auto; margin-right: auto;" />

</br>

# Table of Contents
- [Key Features](#key-features-)
- [Get Started](#get-started-)
  - [Download the Dataset](#download-the-dataset)
  - [Dataset Structure](#dataset-structure)
- [License and Citation](#license-and-citation)

# 👨‍🏫  Get Started    
## Download the Dataset 
To download the full dataset, use the following code. If you encounter any issues, refer to the official Hugging Face documentation.

```bash
# Ensure git-lfs is installed (https://git-lfs.com)
git lfs install
# When prompted for a password, use an access token with write permissions.
# Generate one in your settings: https://huggingface.co/settings/tokens
git clone https://huggingface.co/datasets/InternRobotics/MotionMillion
# To clone without large files (only their pointers)
GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/datasets/InternRobotics/MotionMillion
```

If you only need to download a specific dataset (e.g., `MotionGV/folder1.tar.gz`), use the following code:

```bash
# Ensure git-lfs is installed (https://git-lfs.com)
git lfs install
# Initialize an empty Git repository
git init MotionMillion
cd MotionMillion
# Set the remote repository
git remote add origin https://huggingface.co/datasets/InternRobotics/MotionMillion
# Enable sparse-checkout
git sparse-checkout init
# Specify the target folders and files
git sparse-checkout set MotionGV/folder1.tar.gz
# Pull the data
git pull origin main
```

## Dataset Processing
### Folder Hierarchy
```
MotionMillion
|-- motion_272rpr
|   |-- Mirror_MotionGV
|   |   |-- folder0.tar.gz
|   |   |-- folder1.tar.gz
|   |   |-- folder2.tar.gz
|   |   |-- folder3.tar.gz
|   |   |-- folder4.tar.gz
|   |   |-- folder5.tar.gz
|   |   |-- folder6.tar.gz
|   |   |-- folder7.tar.gz
|   |   |-- folder8.tar.gz
|   |   `-- folder9.tar.gz
|   |-- Mirror_MotionLLAMA
|   |   |-- finedance.tar.gz
|   |   |-- fit3d.tar.gz
|   |   |-- hi4d.tar.gz
|   |   |-- humansc3d.tar.gz
|   |   |-- interhuman.tar.gz
|   |   |-- interx.tar.gz
|   |   `-- trumans.tar.gz
|   |-- Mirror_MotionUnion
|   |   |-- 100STYLE_smpl.tar.gz
|   |   |-- CombatMotion_seperate.tar.gz
|   |   |-- EgoBody.tar.gz
|   |   |-- animation.tar.gz
|   |   |-- fitness.tar.gz
|   |   |-- game_motion.tar.gz
|   |   |-- haa500.tar.gz
|   |   |-- humman.tar.gz
|   |   |-- idea400.tar.gz
|   |   |-- kungfu.tar.gz
|   |   |-- music.tar.gz
|   |   `-- perform.tar.gz
|   |-- MotionGV
|   |   |-- folder0.tar.gz
|   |   |-- folder1.tar.gz
|   |   |-- folder2.tar.gz
|   |   |-- folder3.tar.gz
|   |   |-- folder4.tar.gz
|   |   |-- folder5.tar.gz
|   |   |-- folder6.tar.gz
|   |   |-- folder7.tar.gz
|   |   |-- folder8.tar.gz
|   |   `-- folder9.tar.gz
|   |-- MotionLLAMA
|   |   |-- finedance.tar.gz
|   |   |-- fit3d.tar.gz
|   |   |-- hi4d.tar.gz
|   |   |-- humansc3d.tar.gz
|   |   |-- interhuman.tar.gz
|   |   |-- interx.tar.gz
|   |   `-- trumans.tar.gz
|   |-- MotionUnion
|   |   |-- 100STYLE_smpl.tar.gz
|   |   |-- CombatMotion_seperate.tar.gz
|   |   |-- EgoBody.tar.gz
|   |   |-- animation.tar.gz
|   |   |-- fitness.tar.gz
|   |   |-- game_motion.tar.gz
|   |   |-- haa500.tar.gz
|   |   |-- humman.tar.gz
|   |   |-- idea400.tar.gz
|   |   |-- kungfu.tar.gz
|   |   |-- music.tar.gz
|   |   `-- perform.tar.gz
|-- mean_std
│   |-- Mean.npy
│   `-- Std.npy
|-- texts.tar.gz
|-- splits.tar.gz
```

Due to data licensing restrictions, we only provide parts of the processed motion data with 272-representation. Among these, MotionGV contains motions captured by our motion capture algorithm; the remaining data is merged from other datasets. 

Due to copyright constraints, BABEL, AIST and HumanML3D cannot be released directly. We will provide detailed data processing workflows.

### Data Processing Steps

1. For all tar.gz files, use `tar -xzvf x.tar.gz` to extract them.
2. For HumanML3D, please refer to [data_process/HumanML3D](data_process/HumanML3D/README.md).
3. For BABEL, please refer to [data_process/BABEL](data_process/BABEL/README.md).
4. For AIST, please refer to [data_process/AIST](data_process/AIST/README.md).

### Processed Data Hierarchy
```
MotionMillion
|-- motion_272rpr
|   |-- BABEL
|   |-- Mirror_BABEL
|   |-- Mirror_MotionGV
|   |   |-- folder0
|   |   |-- folder1
|   |   |-- folder2
|   |   |-- folder3
|   |   |-- folder4
|   |   |-- folder5
|   |   |-- folder6
|   |   |-- folder7
|   |   |-- folder8
|   |   `-- folder9
|   |-- Mirror_MotionLLAMA
|   |   |-- aist
|   |   |-- finedance
|   |   |-- fit3d
|   |   |-- hi4d
|   |   |-- humansc3d
|   |   |-- interhuman
|   |   |-- interx
|   |   `-- trumans
|   |-- Mirror_MotionUnion
|   |   |-- 100STYLE_smpl
|   |   |-- CombatMotion_seperate
|   |   |-- EgoBody
|   |   |-- animation
|   |   |-- fitness
|   |   |-- game_motion
|   |   |-- haa500
|   |   |-- humanml
|   |   |-- humman
|   |   |-- idea400
|   |   |-- kungfu
|   |   |-- music
|   |   `-- perform
|   |-- Mirror_PhantomDanceDatav1.1
|   |-- MotionGV
|   |   |-- folder0
|   |   |-- folder1
|   |   |-- folder2
|   |   |-- folder3
|   |   |-- folder4
|   |   |-- folder5
|   |   |-- folder6
|   |   |-- folder7
|   |   |-- folder8
|   |   `-- folder9
|   |-- MotionLLAMA
|   |   |-- aist
|   |   |-- finedance
|   |   |-- fit3d
|   |   |-- hi4d
|   |   |-- humansc3d
|   |   |-- interhuman
|   |   |-- interx
|   |   `-- trumans
|   |-- MotionUnion
|   |   |-- 100STYLE_smpl
|   |   |-- CombatMotion_seperate
|   |   |-- EgoBody
|   |   |-- animation
|   |   |-- fitness
|   |   |-- game_motion
|   |   |-- haa500
|   |   |-- humanml
|   |   |-- humman
|   |   |-- idea400
|   |   |-- kungfu
|   |   |-- music
|   |   `-- perform
|   `-- PhantomDanceDatav1.1
|-- texts
|   |-- Mirror_MotionGV
|   |-- Mirror_MotionLLAMA
|   |-- Mirror_MotionUnion
|   |-- MotionGV
|   |-- MotionLLAMA
|   `-- MotionUnion
|-- mean_std
│   |-- Mean.npy
│   `-- Std.npy
|-- split
|   `-- version1
|       |-- t2m_60_300
|       |   |-- all.txt
|       |   |-- test.txt
|       |   |-- train.txt
|       |   `-- val.txt
|       `-- tokenizer_96
|           |-- test.txt
|           |-- train.txt
|           `-- val.txt
```


# License and Citation
All the data and code within this repo are under [CC BY-NC-SA 4.0](https://creativecommons.org/licenses/by-nc-sa/4.0/). Please consider citing our project if it helps your research.

```BibTeX
@article{fan2025go,
  title={Go to Zero: Towards Zero-shot Motion Generation with Million-scale Data},
  author={Fan, Ke and Lu, Shunlin and Dai, Minyue and Yu, Runyi and Xiao, Lixing and Dou, Zhiyang and Dong, Junting and Ma, Lizhuang and Wang, Jingbo},
  journal={arXiv preprint arXiv:2507.07095},
  year={2025}
}
```

In addition, please cite the following literature:
```BibTeX
@article{xiao2025motionstreamer,
  title={MotionStreamer: Streaming Motion Generation via Diffusion-based Autoregressive Model in Causal Latent Space},
  author={Xiao, Lixing and Lu, Shunlin and Pi, Huaijin and Fan, Ke and Pan, Liang and Zhou, Yueer and Feng, Ziyong and Zhou, Xiaowei and Peng, Sida and Wang, Jingbo},
  journal={arXiv preprint arXiv:2503.15451},
  year={2025}
}

@inproceedings{amass,
  title={AMASS: Archive of motion capture as surface shapes},
  author={Mahmood, Naureen and Ghorbani, Nima and Troje, Nikolaus F and Pons-Moll, Gerard and Black, Michael J},
  booktitle={ICCV},
  pages={5442--5451},
  year={2019}
}

@InProceedings{Guo_2022_CVPR,
    author    = {Guo, Chuan and Zou, Shihao and Zuo, Xinxin and Wang, Sen and Ji, Wei and Li, Xingyu and Cheng, Li},
    title     = {Generating Diverse and Natural 3D Human Motions From Text},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2022},
    pages     = {5152-5161}
}

@inproceedings{babel,
  title={BABEL: Bodies, action and behavior with english labels},
  author={Punnakkal, Abhinanda R and Chandrasekaran, Arjun and Athanasiou, Nikos and Quiros-Ramirez, Alejandra and Black, Michael J},
  booktitle={CVPR},
  pages={722--731},
  year={2021}
}

@inproceedings{flag3d,
  title={Flag3d: A 3d fitness activity dataset with language instruction},
  author={Tang, Yansong and Liu, Jinpeng and Liu, Aoyang and Yang, Bin and Dai, Wenxun and Rao, Yongming and Lu, Jiwen and Zhou, Jie and Li, Xiu},
  booktitle={CVPR},
  pages={22106--22117},
  year={2023}
}

@inproceedings{li2023finedance,
  title={FineDance: A Fine-grained Choreography Dataset for 3D Full Body Dance Generation},
  author={Li, Ronghui and Zhao, Junfan and Zhang, Yachao and Su, Mingyang and Ren, Zeping and Zhang, Han and Tang, Yansong and Li, Xiu},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
  pages={10234--10243},
  year={2023}
}

@article{motionx,
  title={Motion-x: A large-scale 3d expressive whole-body human motion dataset},
  author={Lin, Jing and Zeng, Ailing and Lu, Shunlin and Cai, Yuanhao and Zhang, Ruimao and Wang, Haoqian and Zhang, Lei},
  journal={NeurIPS},
  year={2024}
}

@article{liang2024intergen,
  title={Intergen: Diffusion-based multi-human motion generation under complex interactions},
  author={Liang, Han and Zhang, Wenqian and Li, Wenxuan and Yu, Jingyi and Xu, Lan},
  journal={International Journal of Computer Vision},
  volume={132},
  number={9},
  pages={3463--3483},
  year={2024},
  publisher={Springer}
}

@inproceedings{interx,
  title={Inter-x: Towards versatile human-human interaction analysis},
  author={Xu, Liang and Lv, Xintao and Yan, Yichao and Jin, Xin and Wu, Shuwen and Xu, Congsheng and Liu, Yifan and Zhou, Yizhou and Rao, Fengyun and Sheng, Xingdong and others},
  booktitle={CVPR},
  pages={22260--22271},
  year={2024}
}

@inproceedings{aist-dance-db,
    author = {Shuhei Tsuchida and Satoru Fukayama and Masahiro Hamasaki and Masataka Goto}, 
    title = {AIST Dance Video Database: Multi-genre, Multi-dancer, and Multi-camera Database for Dance Information Processing}, 
    booktitle = {Proceedings of the 20th International Society for Music Information Retrieval Conference, {ISMIR} 2019},
    address = {Delft, Netherlands}, 
    year = 2019, 
    month = nov}

@inproceedings{jiang2024scaling,
  title={Scaling up dynamic human-scene interaction modeling},
  author={Jiang, Nan and Zhang, Zhiyuan and Li, Hongjie and Ma, Xiaoxuan and Wang, Zan and Chen, Yixin and Liu, Tengyu and Zhu, Yixin and Huang, Siyuan},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={1737--1747},
  year={2024}
}

@inproceedings{yin2023hi4d,
  title={Hi4d: 4d instance segmentation of close human interaction},
  author={Yin, Yifei and Guo, Chen and Kaufmann, Manuel and Zarate, Juan Jose and Song, Jie and Hilliges, Otmar},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={17016--17027},
  year={2023}
}

@inproceedings{fieraru2021learning,
  title={Learning complex 3D human self-contact},
  author={Fieraru, Mihai and Zanfir, Mihai and Oneata, Elisabeta and Popa, Alin-Ionut and Olaru, Vlad and Sminchisescu, Cristian},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  volume={35},
  number={2},
  pages={1343--1351},
  year={2021}
}

@inproceedings{danceformer,
  title={Danceformer: Music conditioned 3d dance generation with parametric motion transformer},
  author={Li, Buyu and Zhao, Yongchi and Zhelun, Shi and Sheng, Lu},
  booktitle={AAAI},
  pages={1272--1279},
  year={2022}
}

```

**Special Notes**
- We would like to express our gratitude to the authors of [FineDance](https://github.com/li-ronghui/FineDance) for granting permission to directly open-source the preprocessed motion data. It is important to note that when generating the 272-dimensional motion representation, we utilized the SMPL-X data provided in [MotionLLAMA](https://github.com/ZeyuLing/MotionLLaMA) with all beta values set to 0, which may differ from the original FineDance data.
- If you intend to use the merged data (excluding MotionGV), please strictly adhere to their respective licenses.