file_id string | bundle string | category string | motion_name string | actor_names string | file_path string | extension string | size_bytes int64 | preview_video_path string | frame_count int64 | frame_rate_Hz float64 | duration_seconds float64 | description string | tags string | preview_video video |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
04991 | premium | badminton | premium-badminton-ralley1-NA | NA | premium/badminton/premium-badminton-ralley1-NA/04991-premium-badminton-ralley1-NA.c3d | c3d | 2,683,664 | premium/badminton/premium-badminton-ralley1-NA/04994-premium-badminton-ralley1-NA.mov | 2,937 | 179.910049 | 16.3 | A badminton sequence featuring a rally sequence, captured as variation 1. | badminton;sports;racket-action;full-body;rally;variation-1;motion-capture | Not supported with pagination yet |
Eyes, JAPAN MoCap Dataset (Raw)
Professional optical motion-capture data, fully owned and copyrighted by Eyes, JAPAN, released free for non-commercial research and education. Commercial use requires a separate paid license.
4,139 clips / 10,255 files, 45 action categories, 57 subjects, delivered in multiple industry-standard formats.
- License: CC BY-NC 4.0 — academic/non-commercial use is free.
- Commercial license: purchase required — contact
info@nowhere.co.jp. - SMPL-X version: a processed version of this data fitted to SMPL-X body models is available at
eyesjapan/eyesjapan-mocap-smplx(different license — see that card).
Dataset structure
├── README.md
├── metadata_clips.csv # 1 row per clip (recommended entry point)
├── metadata_files.csv # 1 row per file
└── data/
├── free/ # 2,960 clips
│ └── <category>/<clip_id>/ e.g. free-accident-dodge-azumi/
│ ├── 00003-....bip
│ ├── 00004-....bvh
│ ├── 00005-....c3d
│ └── 00006-....fbx
└── premium/ # 1,179 clips
└── <category>/<clip_id>/...
Every file has a unique zero-padded 5-digit ID; names follow
tier-category-action-subject.
Cleanup before upload: remove the stray
...catwalk-NA.mnmand...hiphop-combo3-umeda.db(a marker-name-map and a Thumbs.db-type file).
File formats
| Ext | Count | Description |
|---|---|---|
.c3d |
3,963 | Raw optical marker data (biomechanics standard) |
.bip |
1,604 | 3ds Max Character Studio Biped motion |
.csm |
1,067 | Character Studio marker file |
.v / .vsk |
886 / 255 | Vicon raw capture / skeleton calibration |
.bvh |
788 | Skeletal animation (hierarchy + rotations) |
.fbx |
690 | Autodesk FBX skeletal animation |
.mov / .m4v |
886 / 116 | Reference video of the capture session |
metadata_clips.csv columns
clip_id, tier (free/premium), category (45 categories), action
(1,624 labels), subject (57 subjects), formats (e.g. bip,bvh,c3d,fbx),
n_files, has_video.
Usage
from datasets import load_dataset
from huggingface_hub import hf_hub_download
index = load_dataset("eyesjapan/eyesjapan-mocap", split="train")
walks = index.filter(lambda x: x["category"] == "walk" and "bvh" in x["formats"])
path = hf_hub_download(
repo_id="eyesjapan/eyesjapan-mocap", repo_type="dataset",
filename="data/free/accident/free-accident-dodge-azumi/00004-free-accident-dodge-azumi.bvh",
)
Data collection & consent
Licensing
- ✅ Academic research, education, personal projects — free under CC BY-NC 4.0
- ✅ Modification & redistribution — with attribution, same non-commercial terms
- ❌ Any commercial use without a separate license from Eyes, JAPAN
- Commercial licensing:
info@nowhere.co.jp
Citation
@misc{eyesjapan_mocap_2026,
title = {Eyes, JAPAN MoCap Dataset (Raw)},
author = {{Eyes, JAPAN}},
year = {2026},
url = {https://huggingface.co/datasets/eyesjapan/eyesjapan-mocap}
}
Takedown
If any content infringes your rights, contact info@nowhere.co.jp and we
will address it promptly.
- Downloads last month
- 9