HoloAssist: an Egocentric Human Interaction Dataset for Interactive AI Assistants in the Real World
Paper • 2309.17024 • Published
Search is not available for this dataset
image imagewidth (px) 896 896 |
|---|
End of preview. Expand in Data Studio
HoloAssist visualization samples
96 paired RGB / colored skeleton overlays from four official HoloAssist recordings. Source authors: Xin Wang, Taein Kwon and the HoloAssist team. Source: https://holoassist.github.io/ ; https://arxiv.org/abs/2309.17024 . Original dataset license: CDLA-Permissive 2.0, included in this repository.
Hand poses are the released HoloLens 2 automatic 3D tracking output, projected with official camera calibration and pose. Not manually annotated RGB 2D points, not marker motion capture, and not MANO. No new pose inference, refitting, smoothing or manual alignment. RGB and overlays are split from previously generated Q95 JPEG pairs and stored as lossless WebP. Invalid / untracked points are omitted. See holoassist_provenance_20261104.json for recordings, frame numbers, timestamps and projection. Sampling favors frames with a tracked hand; not a dataset-wide quality audit. This is an unofficial visualization sample, not a full training release.
- Downloads last month
- -