All datasets

Dataset program

Give household robots the messy, varied demonstrations that staged labs cannot provide.

Human demonstration data from real household tasks including food preparation, cutting, pouring, folding, cleaning, and deformable-object manipulation.

Video context. A household task sequence showing bimanual food preparation and object manipulation for embodied AI training.

Built for model teams

Useful coverage, documented clearly.

Every program is scoped around the task decisions your team needs to make, then delivered with the context needed for review, training, and audit.

Food preparation and kitchen manipulation
Folding, sorting, pouring, and cleaning
Deformable objects and changing object states
First-person hand-object interaction learning

Capture specification

EnvironmentsReal homes, kitchens, tables, counters, and storage areas
TasksPeeling, cutting, folding, pouring, cleaning, and sorting
ViewsEgocentric stereo capture with optional exocentric context
DeliveryTime-aligned video, task labels, object state, hand labels, and QA

Common questions

Before you scale the collection.

How do you handle privacy in household capture?

Consent, anonymization review, and sensitive-background checks are included before buyer delivery.

Can we request specific household objects?

Yes. Object sets, task phases, and environmental constraints are defined in the capture brief.

Continue reading

Read the related robotics data guide

Start with a focused evaluation.

Share the task, objects, environment, target hours, and schema. We will recommend the smallest useful sample.

Request a data brief