← Robotic Datasets

Fiund / Dataset solution

Robot manipulation datasets

Specify tasks, end effectors, actions, and observable outcomes for manipulation learning.

The living-room recording set with brown couches, a round table, and microphones.
Original studio photography · Podcast Studio AZSource & media notes ↗

Manipulation data needs more than a video of an object moving. A useful brief connects the robot platform, controller, observations, and action meaning to the task. Fiund can review sourcing requirements for a particular project; the site does not imply that compatible robot trajectories or approved delivery quantities are already available.

Define the physical task

Describe the initial scene, target outcome, objects, and permitted variation. Opening a drawer, placing an item, and assembling parts involve different contact conditions and success evidence. State whether partial completion, retries, or assistance belong in the requested collection, and whether setup and reset behavior should remain in the episode.

Name the embodiment and controller

Identify the arm configuration, gripper or tool, relevant sensors, and control interface. Action values are only interpretable with their units, frame conventions, and execution meaning. Ask whether records contain commands, measured state, or both. An exported array should not be assumed compatible with the buyer’s robot simply because its dimensions look familiar.

Connect observations to actions

Specify required camera streams, robot state fields, timestamps, and the relationship between an observation and the command associated with it. Review movement around contact and gripper changes. Timing errors can produce a plausible-looking replay while pairing the learner’s input with an action from the wrong part of the task.

Inspect task-level evidence

Request representative episodes with beginning states, execution, and final outcomes. Check command semantics and missing data alongside the media. Review successful and unsuccessful attempts according to explicit definitions. Human demonstration video can provide contextual behavior, but a separate robot action requirement must be supported by actual control records. Preserve object and setup references when repeated attempts are recorded from the same physical arrangement.

Include in your inquiry

  • Robot, end effector, and controller
  • Task setup and completion evidence
  • Observation and action field definitions
  • Episode timing and review criteria

Your project, in focus

Send the task, embodiment, and signals your project requires.

Speak directly with Jaeden Schafer at Fiund about the source material, specifications, and review your project needs.

Source fit, permissions, samples, pricing, and delivery are reviewed for the individual project.

Start with these details

  1. Task and target embodiment
  2. Required observation and action signals
  3. Episode counts and success criteria
  4. Sensor timing and annotation requirements
  5. Source permissions and intended use
Build your project brief ↗

Original studio and recording photography. See the media notes for source context.