![](data:image/svg+xml,<svg display="block" role="presentation" viewBox="0 0 48 48" xmlns="http://www.w3.org/2000/svg"><path d="M 11.308 13.415 L 0 2.108 L 2.108 0 L 11.308 9.2 L 20.508 0 L 22.615 2.108 Z" fill="rgb(0, 0, 0)" height="13.41500000000002px" id="usTXbR14m" transform="translate(12.693 16.693)" width="22.61500000000001px"/></svg>)
Real-world data for Physical AI
Human demonstrations and robot data, collected, annotated and evaluated for your models.
Trusted by Leading AI Teams
UMI
Human demonstrations with handheld grippers, collected in diverse real-world environments.
Tracking
SLAM-based or hardware-based tracking
Cameras
Three synchronized phones or custom cameras
Coordinates
One global, gravity-aligned frame for all cameras where available
Gripper
Custom shapes and looks
Gripper opening
Marker-based or sensor-based state
Sync
Through our capture app or a single hardware clock
QA
Automatic checks on every episode
Labels
Task and action annotations on request
Sample size
12 categories · 80 hours · 1,000 episodes
Egocentric video
First-person video of everyday household tasks, collected in real environments.
HomER
Open
Capture
Head-mounted, first-person video
Hands
Both hands visible at least 95% of the time
Footage
No third-person video or reused clips
QA
Automatic quality checks on every episode
Sample size
17 categories · 1.7 hours · 63 videos
Collect bespoke egocentric data from 200,000+ contributors in 100+ countries with our egocentric self-service.
Robot teleoperation data
Robot data captured through VR teleoperation in our robotics lab.
Teleop
Robots
Single-arm or bimanual manipulator
Tracking
Joint states, end-effector pose and actions (TCP velocities)
Cameras
Three synchronized views with depth
Gripper
Custom shapes and looks
Gripper opening
Sensor-based state
Sync
Through a single hardware clock
QA
Both manual and automatic checks
Lables
Task and action annotations
Sample size
20 categories · 40 hours · 5,000 episodes
Collect with Toloka
Partner with Toloka to collect data using your operators, robots, or capture setup.
Custom data programs
Custom collection built around your requirements, from the tasks and capture setup to the annotation schema and final delivery.
Capture
Define tasks, participants, hardware and environments. Collect in our lab or across distributed sites.
01
Quality assurance
Check recording quality, timing and sensor alignment. Route flagged sequences to expert review.
Acceptance criteria: agreed with you before collection starts.
Privacy: blurring, masking, access and retention set to your requirements.
02
Annotate and deliver
Add frame labels, temporal boundaries and task success assessments.
03
Your grippers
Customize the shape and look of UMI and robot grippers.
Your labels
Temporal annotations of tasks and actions.
Your format
Delivery as MCAP, LeRobot, MP4 or custom schema.
What should your robot learn?
Tell us what you're building. We'll help plan the data.