Delta-Intelligence/OmniEgo

D1 Headset Egocentric Whole-body Dataset

D1 is a headset multi-camera human motion dataset for humanoid intelligence, embodied AI, whole-body motion understanding, and imitation learning.

Overview

The D1 dataset is exported from the D1 headset multi-camera human motion capture system developed by Delta Intelligence. Each recorded episode contains synchronized multi-view video streams and whole-body skeleton and headset pose data.

The dataset supports research and development in:

• Human behavior observation and understanding
• Human detection, keypoint detection, and pose estimation
• Headset device trajectory and motion analysis
• Egocentric visual observation and synchronization checking
• Whole-body motion modeling for humanoid robots
• Imitation learning, behavior cloning, and embodied intelligence model training
• Visual-inertial odometry and SLAM-related observation analysis

Unlike purely third-person datasets, D1 combines egocentric RGB and fisheye views with global whole-body skeleton and headset pose states, making it suitable for learning from human demonstrations in realistic environments.

Dataset Summary

The current public release contains 552 episodes, 133,246 master frames, and 4,441.533 seconds of synchronized data.

Task	Episodes	Master frames
dish-washer	89	30,687
fridge	13	6,782
move-boxes	2	430
oven	268	48,228
washing	180	47,119

Headset Device

The D1 headset is designed for full-body human data acquisition from an egocentric perspective.

Parameter	Description
Processor	Rockchip RK3588-class processor
Memory / Storage	4GB + 32GB / 8GB + 64GB configurations
Connectivity	Dual-band Wi-Fi, Bluetooth 5.3, Type-C interface
Audio	Dual microphones and speaker-supported audio collection
Battery	7.4V 1000mAh
Charging	Dedicated charging base
Tracking Accuracy	Full-body system spatial positioning accuracy within approximately 2 cm
Data Types	RGB image, IMU data, whole-body skeleton data, microphone audio, spatial pose trajectory

Dataset Structure

The dataset is distributed as MCAP files and matching Parquet files for the dataset viewer.

.
├── mcap/
│   └── <task>/<episode>.mcap       # One MCAP file per episode
├── data/
│   └── <task>/<episode>.parquet    # Frame rows with embedded JPEG images
├── example.parquet                 # Compact viewer preview
└── manifest.jsonl                  # Counts, dimensions, and file sizes

Each episode provides synchronized visual observations, headset pose and motion state, and whole-body skeleton data when available.

Video Streams

The release contains a main RGB camera, two side RGB cameras, and two tracking cameras. The main camera is 3840 x 2160 at 30 FPS, the side cameras are 1280 x 720 at 15 FPS, and the tracking cameras are 848 x 800 at 30 FPS.

Typical uses include human action visualization, pose estimation, multi-view visibility checking, headset motion analysis, and synchronization validation.

Skeleton and State Data

The skeleton data uses a 24-joint SMPL-style body layout with frame-aligned 3D joint positions. Headset state data includes translation, rotation quaternion, linear velocity, angular velocity, and acceleration.

These fields support whole-body motion visualization, video-skeleton synchronization checks, motion prediction, imitation learning, and humanoid motion retargeting.

Privacy and Data Quality

Only the main camera uses a fixed circular blur region fitted to the wearer's head in each episode. Side and tracking camera images are restored from the source export. JPEG payloads are shared between the MCAP and Parquet representations.

Each episode was read back after conversion. Validation covers frame counts, stream-to-master mappings, timestamps, MCAP CRCs, and matching image payload hashes between the MCAP and Parquet files. Missing source values remain missing; no pose or skeleton values are fabricated.

Loading

from datasets import load_dataset

ds = load_dataset(
    "parquet",
    data_files="hf://datasets/Delta-Intelligence/OmniEgo/data/*/*.parquet",
    split="train",
    streaming=True,
)
row = next(iter(ds))

Contact

For questions, feedback, or collaboration requests:

• Email: sales@deltai.com
• Website: https://www.deltai.com

Acknowledgement

This dataset is collected using the Delta Intelligence D1 data acquisition system, designed for humanoid intelligence, embodied AI, and whole-body human motion data collection.