ego-multimodal — MoWare IMU Motion Capture + GMR

Interactive Rerun visualizations for mocap, video, segmentation, and General Motion Retargeting

Mocap

Video with activity annotations, hand-proximal segmentation masks, and full-body IMU mocap avatar.

video segmentation annotations mocap avatar

2.9 GB

Open in Rerun Viewer Download .rrd

Finger Dexterity

3D pinhole camera projection with head tracking, full mocap skeleton, and close-up views of both hands.

pinhole camera head tracking hand close-ups

1.8 GB

Open in Rerun Viewer Download .rrd

General Motion Retargeting

Human motion retargeted to 3 humanoid robots via GMR transfer functions: Unitree G1, Fourier N1, and Booster T1.

Unitree G1 Fourier N1 Booster T1 GMR

10 GB

Open in Rerun Viewer Download .rrd
This file is 10 GB. The Rerun web viewer may struggle with files this large. For best results, download the .rrd and open locally with rerun gmr_robots.rrd.

Programmatic Access

Load the structured data (quaternions, trajectories, annotations) directly via the datasets library:

from datasets import load_dataset # Load motion capture quaternion data (78K rows) ds = load_dataset("phi-9/ego-multimodal", "mocap") # Load activity annotations ann = load_dataset("phi-9/ego-multimodal", "annotations") # Load 3D trajectories head = load_dataset("phi-9/ego-multimodal", "trajectory_head_only") # Download a Rerun visualization from huggingface_hub import hf_hub_download path = hf_hub_download("phi-9/ego-multimodal", "visualizations/video.rrd", repo_type="dataset")

Dataset Contents

Full file listing at phi-9/ego-multimodal

data/mocap_frames.parquet 883 KB
data/calibration_frame.parquet 3.5 KB
data/trajectory_*.parquet (×4) ~9 KB each
data/annotations.parquet 3.2 KB
visualizations/video.rrd 2.9 GB
visualizations/pinhole.rrd 1.8 GB
visualizations/gmr_robots.rrd 10 GB
segmentation/l2.hand_prox_seg.json 1.1 GB
raw/l2.mp4 76 MB
raw/l2_mod_combined.decoded.json 27 MB