Skilled people, recorded at work on our own stereo hardware. What they see, what their hands do, what they feel. Rebuilt in simulation, ready to train on.
Every skill a robot will ever have was a human skill first.
What the cap sees, what the hands do, what they feel. Then the scene as a measurement, as a simulation, and as a robot doing the task.
01captureA cap with two cameras, worn all day by people who know the job.
02handsTwenty one points per hand, every frame. How a grip forms, and how it lets go.
03touchWhere the fingers press, how hard, and for how long.
04reconstructTwo views become depth. The bench becomes a measurement.
05simulateThe same scene, rebuilt where a policy can fail a thousand times for free.
06transferTrained on people. Checked in simulation. Handed to a robot.
Hours collected and sold
Hours a day of collection capacity
Regions: USA, India and Southeast Asia
Our own capture hardware, in the field
faces blurred before deliveryfaces blurred before delivery01 / 10
Real recordings, with the signals a model trains on. Recording, episode, environment, device, operator.
Data is the product. Capture, hands and simulation are what make it compound.
A stereo camera cap people can wear all day.
Two synced cameras under the brim of an ordinary cap, built in house. It records what skilled people see and do, at work, without a studio or a script. Every recording carries a device clock anchor so the two streams and the sensor data line up to the millisecond.
Hand pose and contact, read from the recordings.
We track both hands in every frame and estimate where they touch and how hard they press. That is the part of human skill a video alone does not show, and the part a dexterous robot needs most.
Real scenes rebuilt as simulations.
Stereo depth turns the workbench, the tools and the hands into a measured scene. We rebuild that scene in simulation so a policy can be trained and tested against the same task, thousands of times, before it touches hardware.
The next step: policies back onto robots.
Planned work. Policies trained on human recordings and checked in simulation are transferred to robot hands and arms, and their failures feed back into what we record next.
A recording becomes a measured scene. A measured scene becomes a simulation.
