简体中文 | English
90,000 小时头戴视角第一人称人类活动数据
所有数据在发布前均经过自动去标识化流程和人工复核,对人脸、车牌及其他个人身份信息进行模糊处理。
| Collection | Data Repositories | SKU | Sub-SKU | Format | Planned Duration | Website |
|---|---|---|---|---|---|---|
| EgoSuite-Open100K | EgoStandard CURRENT DATASET EgoPro EgoDemo | EgoStandard | EgoStandEgoStand-body | LeRobot MCAP | 90,000 hours | EgoSuite Website |
EgoSuite-Open100K 是一个分阶段发布的 100,000 小时开放数据集,旨在兼顾规模、多样性与标注质量。数据覆盖 15,000+ 项任务和 15,000+ 个不同采集场景,提供手部与全身位姿;部分数据包含腕部相机视角,并为选定子集免费附赠事件级语义标注。
EgoStandard 是 EgoSuite-Open100K 中规划时长 90,000 小时的头戴视角数据线,将第一人称视频与同步 3D 手部位姿以及可选的全身位姿相结合。数据合集覆盖 15,000+ 项任务和 15,000+ 个不同采集场景;这些是整个发布项目共享的合集级覆盖数据,并非每个仓库的独立计数。LeRobot 与 MCAP 是同一批 EgoStandard Episode 和相同 90,000 小时规划数据的不同表示形式,因此不得相加。
| Sub-SKU | Duration | Standard SKU definition |
|---|---|---|
EgoStand | 80,000 hours | Head-view video + left/right hand pose |
EgoStand-body | 10,000 hours | Head-view video + left/right hand pose + full-body pose |
EgoStand 是面向大规模手部中心活动学习的高容量头戴视角变体。
EgoStand-body 增加全身位姿,用于关联手部交互与全身运动。
两个 EgoStandard Sub-SKU 均不包含腕部视角视频。
标准 EgoStandard 产品线包含左右手位姿,以及 EgoStand-body 中的全身位姿;语义标注不属于标准 SKU 定义。针对本次开源发布,光轮智能免费提供时序、事件级语义标注作为附加内容。
| Format | Recommended use | Location | Episode layout | Documentation |
|---|---|---|---|---|
| LeRobot v3.0-compatible | Training and LeRobot-compatible robotics pipelines | EgoStand/lerobot/, EgoStand-body/lerobot/ | Per-episode Parquet, video, LeRobot metadata, and release annotation files | Sub-SKU README files |
| MCAP | Synchronized multimodal streams and MCAP-compatible workflows | EgoStand/mcap/, EgoStand-body/mcap/ | Per-episode MCAP package and any verified sidecar files | Sub-SKU README files |
EgoStandard/
├── README.md
├── EgoStand/
│ ├── README.md
│ ├── lerobot/
│ ├── mcap/
│ └── manifests/
└── EgoStand-body/
├── README.md
├── lerobot/
├── mcap/
└── manifests/旧版 data/EgoStand/ 目录包含现有受保护数据,因此有意保留且不作修改。所有新增上传必须使用上述 Sub-SKU 优先路径。
| Modality | EgoStand standard SKU | EgoStand-body standard SKU | Included in this open-source release |
|---|---|---|---|
| Head-view egocentric video | Yes | Yes | Yes |
| Left/right 3D hand pose | Yes | Yes | Yes |
| Full-body 3D pose | No | Yes | Body variant only |
| Wrist-view video | No | No | No |
| Semantic annotations | No | No | Yes, complimentary add-on |
| Frame-quality metadata | Delivery-dependent | Delivery-dependent | Included when available |
位姿值采用 float32 类型和世界坐标。手部位置数组 shape 为 (21, 3),手部旋转为 (21, 4),全身位置为 (22, 3),全身旋转为 (22, 4)。四元数值遵循 (qw, qx, qy, qz) 顺序。
LeRobot 与 MCAP 是本次发布数据的不同表示。各格式 manifest 使用相同且稳定的 episode_uuid 和 sub_sku 值来标识对应 Episode。计算数据集规模时,用户不得将 LeRobot 与 MCAP 的行数相加。
通用 manifest 字段如下:
| Field | Description |
|---|---|
episode_uuid | Stable episode identifier shared across formats |
human_case_id | Stable Human Case identifier |
sku | Always EgoStandard |
sub_sku | EgoStand or EgoStand-body |
task_name | Normalized task name |
scene_id | Audited scene identifier |
scene_family | Top-level scene family |
duration_s | Audited usable duration in seconds |
episode_path | Repository-relative path for this format |
format | lerobot-v3.0 or mcap |
schema_version | Format-specific schema/profile version |
has_hand_pose | Whether hand pose is present |
has_body_pose | Whether full-body pose is present |
annotation_version | Internal schema identifier when the complimentary semantic annotation is present |
annotation_distribution | complimentary-open-source-add-on when applicable |
source_revision | Immutable source revision |
sha256 | Integrity checksum |
各格式专属字段记录在对应格式的 README 中。
| Release item | Value |
|---|---|
| Dataset release version | [VERIFY] |
| LeRobot format | v3.0-compatible |
| MCAP version/profile | [VERIFY] |
| MCAP schema | [VERIFY] |
| Converter version | [VERIFY] |
| Semantic annotations | Included as a complimentary release add-on; internal version not exposed |
| Annotation distribution | Complimentary open-source release add-on |
仅下载 LeRobot:
git clone https://ai.gitcode.com/Lightwheel/EgoStandard.git
cd EgoStandard
git lfs pull --include="EgoStand/lerobot/**,EgoStand-body/lerobot/**"仅下载 MCAP 时,将包含路径替换为 EgoStand/mcap/**,EgoStand-body/mcap/**。