xuejf/EgoDemo-v2-Test

EgoDemo

A 20-50 hour ungated trial pack sampled from the formal EgoSuite-Open10K repositories.

Collection · EgoStandard · EgoPro · Interactive Visualizer · 中文

Private v2 test repository. The release target is public and ungated, but the audited Demo manifest, exact duration, selected media, approved license, checksums, and final citation have not been supplied. The Viewer remains disabled until the downloadable sample pack exists.

Trial-pack contract

EgoDemo is a copied sample pack, not additional hours. Every released row must preserve source provenance and set counts_toward_10k=false.

Property	Required value
Access	Public and ungated after approval
Duration	20-50 audited hours
Official 10K contribution	Zero
Source coverage	Selected rows from the six formal configs
Required provenance	Repository, config, clip, revision, checksum
Viewer	Enabled only after the audited manifest is uploaded

Required provenance fields

Field	Type	Meaning
demo_clip_id	string	Demo-local anonymous key
source_repo	string	xuejf/EgoStandard-v2-Test or xuejf/EgoPro-v2-Test
source_config	string	One of the six formal config names
source_clip_id	string	Stable source clip key
source_revision	string	Immutable source revision
counts_toward_10k	bool	Must always be false
scene_family	string	One of the seven environment families
duration_s	float32	Audited usable duration
head_video_ref	string	Demo media reference
wrist_video_ref	string, nullable	Present for sampled Pro rows
hand_pose_3d_ref	string	Hand-pose reference
body_pose_3d_ref	string, nullable	Body-pose reference when sampled
semantics_v7_ref	string, nullable	V7 semantic reference when sampled
sha256	string	Integrity checksum

Hero-SKU samples

The selection plan includes representative EgoFull and EgoProMax rows so users can inspect the highest-supervision schemas. It must not imply that the Demo currently contains those rows before the audited source map is uploaded.

Open EgoSuite-Visualizer for the environment hover interaction and hero-SKU comparison.

Target loading interface

from datasets import load_dataset

demo = load_dataset(
    "xuejf/EgoDemo-v2-Test",
    split="train",
    streaming=True,
)

sample = next(iter(demo))
assert sample["counts_toward_10k"] is False
print(sample["source_repo"], sample["source_config"], sample["source_clip_id"])

License and responsible use

The requested cc-by-nc-sa-4.0 identifier is not declared because no approved dataset license was supplied. The Demo must use terms that are compatible with every copied source row and with participant consent.

Users must not identify participants, reconstruct sensitive locations, or use the data for surveillance, profiling, or harmful applications.