Humanoid robot training data
A humanoid that makes a bed or loads a dishwasher has to learn it from people doing the same job. We film that work from the worker's own point of view, in the rooms where it happens.
Tasks are recorded by trained operators in real homes, hotels and kitchens in five countries, so one dataset covers different tools, rooms and habits.
Every task is framed so hands, objects and contact stay visible. That is what pose and action models need.
On request we add task labels, action timestamps and hand pose data.
Scripted slips, drops and retries, each one time-stamped. The cases natural footage rarely contains.
The worker's view and the room's view, for teams training world models.
Body tracking and tactile gloves, scoped and priced per project.
The same protocol recorded in every location.
Tasks, device, metadata and acceptance criteria, agreed in writing.
A small trained group records a first batch. You judge real footage.
More operators and more sites, supervised on location.
Accepted clips, metadata and the consent trail, in your structure.
Bed making, dishwashing, food prep, surface cleaning, laundry and linen, loading and stocking. Send your own task list and we find people who do that work for a living.
Yes. Operators use the device you specify or ship to us, including handheld grippers.
The pilot crew is small on purpose. Once the data is right, we add operators through our hospitality networks, from tens to thousands, with team leads on site.
You choose. Footage can be exclusive to you, or non-exclusive at a lower price. It is agreed per project and written into the contract.
Send the tasks and the spec. We reply with a pilot proposal covering scope, timeline and price.