Openfloor

The company turns customer-specific tasks, recorded on iPhones by invited contributors, into annotated datasets designed to support robot learning. Robotics labs can define a task and its requirements, have contributors capture the requested activity, review annotations and data quality, and receive recording episodes with documentation.

The data may include original RGB video, frame timestamps, camera pose and intrinsics, and, depending on the setup and processing, depth and confidence information, reviewable hand estimates, and optional phone motion measurements. openfloor keeps original measurements and derived estimates distinct, with their timing, availability, and provenance. Signals, formats, and interpretation can be discussed for each collection.

Contributors can add actions, objects, and key moments to recordings, submit them for review, and follow feedback in a workspace. Participation is by invitation; applying does not guarantee access or paid work. Labs can contact openfloor to discuss a custom pilot, including task definition, capture setup, annotations, and delivery.