About this service
ManuData collects first-person video of people performing real tasks on factory floors in India. This service is for robotics and Physical AI teams that need human demonstration data for robot training. Projects can cover picking, packing, sorting, assembly, quality checks, lacing, bonding, and other factory tasks. ManuData uses cap-mounted cameras to record the worker's view while the task is performed. The data can include 1080p video, hand tracking, camera movement, object masks, action labels, task descriptions, and timestamps. Each project is set up around the buyer's task list and required output. The final data can be delivered in RLDS or LeRobot format for model training and evaluation.
Collection methods
Data types
Industries
Geographic coverage
Languages
Delivery formats
Typical deliverables
First-person factory video, extracted images, hand keypoints, camera poses, object masks, action labels, task descriptions, timestamps, metadata, and a dataset manifest. Final data is delivered in RLDS or LeRobot format as agreed for the project.
Quality assurance
ManuData processes captures through camera tracking, hand tracking, object segmentation, world anchoring, action labeling, and final RLDS or LeRobot export.

