README.md

---
pretty_name: "EgoSuite-Open10K · EgoStandard"
language:
- en
- zh
tags:
- robotics
- embodied-ai
- physical-ai
- egocentric-video
- human-demonstration
- hand-pose
- body-pose
- semantic-annotation
- multimodal
- streaming
viewer: false
configs:
- config_name: EgoStandard
  data_files:
  - split: train
    path: "manifests/EgoStandard/*.parquet"
- config_name: EgoStand-motion
  data_files:
  - split: train
    path: "manifests/EgoStand-motion/*.parquet"
- config_name: EgoFull
  data_files:
  - split: train
    path: "manifests/EgoFull/*.parquet"
---

<div align="center">

<img src="./assets/hero.png" width="100%" alt="EgoSuite-Open10K — real-world egocentric human data by Lightwheel" />

# EgoStandard

### 9,000 hours of head-view human activity with progressive pose and semantic supervision.

**Part of EgoSuite-Open10K · 3 configs · 7 scene families · clip-aligned annotation layers**

[Collection](https://huggingface.co/collections/xuejf/egosuite-open10k) · [EgoPro](https://huggingface.co/datasets/xuejf/EgoPro) · [Open Demo](https://huggingface.co/datasets/xuejf/EgoDemo) · [中文](./README_zh.md) · [EgoSuite](https://lightwheel.ai/egosuite)

</div>

> **Public landing card · manual gated access.** This repository is now publicly discoverable, while access requests use manual review. The Dataset Viewer remains disabled until the audited Parquet manifests, media shards, approved license, and final citation are uploaded. This public card does not itself grant rights to unpublished training data.

## EgoSuite-Open10K at a glance

EgoSuite-Open10K is Lightwheel's 10,000-hour egocentric human-data release for embodied AI and world-model research. It is organized as one Hugging Face Collection and three repositories so users can enter through any repo while retaining a complete view of the release.

<img src="./assets/repository-map.svg" width="100%" alt="EgoSuite-Open10K collection and repository architecture" />

| Repository | Role | Configs | Formal hours | Access at release |
|---|---|---|---:|---|
| **EgoStandard** | Head-view series | EgoStandard / EgoStand-motion / EgoFull | **9,000** | Manual gated |
| [EgoPro](https://huggingface.co/datasets/xuejf/EgoPro) | Head + wrist series | EgoPro / EgoPro-motion / EgoProMax | **1,000** | Manual gated |
| [EgoDemo](https://huggingface.co/datasets/xuejf/EgoDemo) | Open duplicate trial pack | EgoDemo | 20–50, excluded from total | Public, ungated |

> The official 10,000-hour total is `9,000 h + 1,000 h`. EgoDemo duplicates selected source clips and is never added to that total.

## Three configs in this repository

<img src="./assets/config-ladder.svg" width="100%" alt="EgoStandard configuration ladder" />

| Config | Format | Hours | Share of this repo |
|---|---|---:|---:|
| **EgoStandard** | Head video + 3D hand pose | **8,400** | 93.33% |
| **EgoStand-motion** | Head video + 3D hand pose + 3D body pose | **500** | 5.56% |
| **EgoFull** | Head video + 3D hand pose + 3D body pose + V7 semantics | **100** | 1.11% |
|  | **Repository total** | **9,000** | **100%** |

### Why configs instead of three repositories?

The capture family stays together while supervision depth is exposed as a first-class Hugging Face config. Each manifest row uses a stable `clip_id`; media and annotation objects are referenced by that key. Body pose and V7 semantics are additive layers on the clip—not duplicated videos created for a higher tier.

```text
clip_id
├── head_video_ref
├── hand_pose_3d_ref
├── body_pose_3d_ref        # motion and full tiers
└── semantics_v7_ref        # full tier
```

## Load a config

Choose a config explicitly so the loaded schema matches the intended SKU:

```python
from datasets import load_dataset

standard = load_dataset(
    "LightwheelAI/EgoStandard",
    "EgoStandard",
    split="train",
    streaming=True,
)

motion = load_dataset(
    "LightwheelAI/EgoStandard",
    "EgoStand-motion",
    split="train",
    streaming=True,
)

full = load_dataset(
    "LightwheelAI/EgoStandard",
    "EgoFull",
    split="train",
    streaming=True,
)
```

The lightweight Parquet manifests are the config entry points. Large video and annotation objects remain versioned once in their canonical paths and are resolved through the reference columns.

## Common manifest contract

| Field | Type | Meaning |
|---|---|---|
| `clip_id` | string | Globally unique, anonymous clip key |
| `config_name` | string | Exact HF config / SKU name |
| `scene_family` | string | One of seven top-level scene families |
| `task_name` | string | Normalized task name |
| `duration_s` | float32 | Audited usable duration in seconds |
| `head_video_ref` | string | Canonical head-view media reference |
| `hand_pose_3d_ref` | string | Frame-aligned 3D hand-pose reference |
| `body_pose_3d_ref` | string, nullable | Frame-aligned 3D body-pose reference |
| `semantics_v7_ref` | string, nullable | Frame-aligned V7 semantic reference |
| `release_revision` | string | Immutable release revision |
| `sha256` | string | Integrity checksum for the row's primary media object |

Config-specific required and nullable fields are defined in [`metadata/schema.json`](./metadata/schema.json).

## Repository layout

```text
EgoStandard/
├── README.md
├── README_zh.md
├── assets/
├── manifests/
│   ├── EgoStandard/part-*.parquet
│   ├── EgoStand-motion/part-*.parquet
│   └── EgoFull/part-*.parquet
├── media/head/<shard>.tar
├── annotations/hand_pose_3d/<shard>.tar
├── annotations/body_pose_3d/<shard>.tar
├── annotations/semantics_v7/<shard>.tar
└── metadata/
    ├── sku_catalog.json
    ├── schema.json
    ├── statistics.json
    └── checksums.sha256
```

## Scene coverage

<img src="./assets/seven-scenarios.png" width="100%" alt="Seven scene families covered by EgoSuite-Open10K" />

The release spans seven top-level families: **Home, Hospitality, Retail, Sports, Logistics, Office, and Industry**. Final per-scene hours must be generated from the audited release manifest rather than estimated in the Dataset Card.

## Access and responsible use

This is a formal-release repository. This public repository uses **manual gated access**. Access approval does not override the dataset license or use restrictions published with the release.

Users must not attempt to identify participants, reconstruct sensitive locations, or use the data for surveillance, profiling, or harmful applications. The release documentation must stay version-aligned with participant authorization, anonymization, privacy review, quality-control records, and the audited manifest.

## License, citation, and contact

The approved dataset license and citation author list are still pending. Public page visibility and manual access review do not grant rights to unpublished training data.

For product information, visit [Lightwheel EgoSuite](https://lightwheel.ai/egosuite). For release or collaboration inquiries, use the official [Lightwheel contact form](https://lightwheel.ai/contact).

```bibtex
@dataset{lightwheel_egosuite_open10k_egostandard_2026,
  author    = {{Lightwheel}},
  title     = {EgoSuite-Open10K: EgoStandard},
  year      = {2026},
  publisher = {Hugging Face},
  url       = {https://huggingface.co/datasets/xuejf/EgoStandard}
}
```

> Citation metadata is a release-candidate template until the approved author list and publication record are locked.