EmbodiedData
Start here →
Datasets/Food preparation/WILD-Mani Kitchen — Bimanual Human Motion
✓ Source card reviewedSee source termsVersion: Parquet + video

WILD-Mani Kitchen — Bimanual Human Motion

102 trimmed bimanual human demonstrations in real kitchens, with Meta Quest 3 skeletal hand tracking, hierarchical action labels, and language-oriented annotations at 30 Hz.

O
Published by OOJUWILD-Mani · Source updated 21 Feb 2026 · Reviewed 17 Jul 2026
View dataset files ↗2.68 GB · Parquet + videoDownloads and licence terms are handled by the original publisher on Hugging Face.
Human demonstration dataset

This dataset may support visual pretraining, task understanding, action recognition, or affordance learning. It does not contain robot action trajectories required for direct policy training.

START HERE

This dataset, explained simply

Intermediate

People perform kitchen activities while a headset records video and the motion of both hands.

WHAT CAN I TRAIN?✓ Action-recognition models✓ Hand-motion understanding✓ Visual pretraining
WHAT DO I NEED?• 2.68 GB storage• Video-processing basics• Not suitable for direct robot policy training
SUITABILITY× Video and human motion only✓ Real-world recordings• Check source for language labels
New to the terminology?Episode = one complete attemptObservation = what the robot sees and sensesAction = the command sent to the robot

VERIFICATION RECORD

What EmbodiedData checked

Source statusAvailable at publisherLast checked17 July 2026Dataset versionParquet + videoTraining instructionsNot tested by EmbodiedData

Verification confirms the linked metadata at the stated date. It is not a guarantee that upstream files, dependencies or licences will remain unchanged.

Dataset previewIllustrative image—not an original dataset frame
Preview only
PUBLISHER-HOSTED MEDIAOpen genuine dataset samplesOOJU · source repositoryView real samples at source ↗
EXAMPLE OF THE TASKSet the plate and fork in the correct place on the table.
View real samples on Hugging Face ↗

Original media, file structure, and dataset viewer are provided by the publisher.

DATASET OVERVIEW

What the publisher reports

102 trimmed bimanual human demonstrations in real kitchens, with Meta Quest 3 skeletal hand tracking, hierarchical action labels, and language-oriented annotations at 30 Hz.

102Reported episodes
30 HzReported frequency
44mReported duration
Parquet + videoDataset format

KNOWN CONFIGURATION

Robot, observations and reuse

Robot or capture systemNo robot — Meta Quest 3 hand tracking
Reported observationsHead-mounted RGB, 2× hand pose, Action labels
ActionsNo robot actions reported
EnvironmentReal-world human motion · 4 tasks
Licence shown on source cardSee source terms

TRUST & LIMITATIONS

What EmbodiedData has—and has not—checked

Original Hugging Face source linkChecked
Summary against publisher cardReviewed
File integrity and missing valuesNot independently tested
Privacy, safety and legal complianceNot independently tested

Always read the original dataset card, licence, limitations and file documentation before training or redistribution.

SOURCE & CITATION

Use the publisher’s citation

EmbodiedData does not invent a citation or dataset version. Open the original card and copy the citation requested by the publisher.

Open original dataset card ↗