We’re mapping demand for humanoid training data

To support the development of capable, real-world humanoid robots, we’re collecting input on training data needs across industries and applications.

This short survey helps us understand what types of data are required, at what level of quality, and for which use cases. Your response helps inform future datasets and market insights.

Thank you for taking a few minutes to contribute.

Name
Choose what describes your company

Tasks & environments

Which markets should the datasets cover?
Task families you want covered
Select the task types that matter most.
“gloved hands”, “oily parts”, “tight spaces”, “low light”, “PPE”, etc.

Scale & distribution

How much data do you need (rough order of magnitude)?
Preferred delivery method
What data format would you require?
If needed, describe how touch should be tied to video (sensor IDs, coordinate frames, calibration steps, etc.)

Covariate shift

Do you care where the data is collected?
How important is environment diversity?
e.g., multiple sites/countries vs single controlled environment.
Would you require exclusivity for the dataset you purchase?
This will lead to an increased price point for the dataset.

Video specs

Video format required
What resolution is required?
Is lossy compression acceptable?
Is audio needed?

Physical data

Do you need object physical properties?
Force / weight annotations needed?
Would you pay 800 USD pr 10 hours of non exclusive data
Would you pay 5000 USD pr 10 hours of exclusive data