Buckets:

LightwheelAI/EgoPro / README_EgoPro_bucket.md
Deyang4910's picture
|
download
raw
2.52 kB

EgoPro

Part of EgoSuite-Open100K, an open release of 100,000 hours of egocentric human data by Lightwheel, in partnership with Hugging Face.

EgoPro adds a wrist-mounted camera view alongside the head-mounted one, resolving contact and grasp at a range the head view cannot reach.


Looking for the dataset?

Start here:

This bucket holds the bulk data for EgoPro. It's object storage rather than a dataset repository, so there's no dataset card or built-in viewer here.


What's in EgoPro

Camera views Head-mounted and wrist-mounted
Annotations Hand pose, body pose, subtask-level semantics
Formats MCAP in this bucket; LeRobot v3 samples in EgoDemo

Annotation coverage varies by subset. See the dataset card on the Hub for the full mapping.


Accessing the data

pip install -U huggingface_hub
hf auth login

# List what's here
hf buckets list LightwheelAI/EgoPro -h -R

# Sync a directory to local
hf buckets sync hf://buckets/LightwheelAI/EgoPro/EgoProStandard-body ./EgoProStandard-body

# Or copy a single file
hf buckets cp hf://buckets/LightwheelAI/EgoPro/path/to/file ./file

Buckets are S3-compatible, so AWS CLI, boto3 and s5cmd also work. See the S3 compatibility guide for configuration.


Privacy

All recordings were processed through an automated de-identification pipeline with human verification, blurring faces, license plates and other personally identifiable information before release.

All participants provided informed consent covering data collection, annotation, controlled release, and the uses permitted under the dataset license.


License

Open for academic research and commercial training, subject to the dataset license and access terms.


Contact

Questions about the data: open a discussion on the EgoDemo repository.

Everything else: contact@lightwheel.ai

Xet Storage Details

Size:
2.52 kB
·
Xet hash:
b43b8b8f11fa04a01d618af79a6c14c744b24ef5edccbbc16a9bb993f29d8192

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.