KB Quality grade is Kinetic Blocks' own grade, A+ to F, measured from the files across four pillars: Technical, Temporal integrity, Content, Completeness.
KB Quality grade is Kinetic Blocks' own grade, A+ to F, measured from the files across four pillars: Technical, Temporal integrity, Content, Completeness. How grading works
What you get
Measured hours
0.2 h
Measured by Kinetic Blocks from the files, not declared by the seller.
You pay Kinetic Blocks, never the seller. You get 5 business days to check the files against the dataset card. Reject for a material mismatch and, once we uphold it, you are refunded in full. How payment works
What we checked against the files
3 of 6 claims verified against the files
3 verified
3 can't confirm
0 doesn't match
Dataset card check
✓verified
IMU timestamps check
✓verified
IMU values check
✓verified
Modality Audio
Declared by the seller, not yet verified.
?can't confirm
Total duration check
Declared by the seller, not yet verified.
?can't confirm
Total Size
Declared by the seller, not yet verified.
?can't confirm
KB Quality grade A, by pillar
A+
A
A-
B+
B
B-
C+
C
C-
D
F
Technical quality30%
99
Temporal integrity25%
100
Content usefulness30%
83
Completeness15%
100
Dead time is mapped, not cut.
100% active task footage (0.2 h of 0.2 h)
Violet is task action, grey is idle, measured by Kinetic Blocks in every clip. Idle stretches lower the grade; they are never cut from the files.
Scores out of 100. Every seller is graded on the same versioned weights. How grading works
Real-world UMI gripper demonstrations for robotic manipulation, featuring RGB video, 6-DoF pose, gripper actions and diverse object interactions.
Overview
Real-world UMI handheld-gripper manipulation demonstrations designed for robotics, Vision-Language-Action (VLA) models, imitation learning, and Physical AI.
The dataset targets diverse object interactions, including grasping, pick-and-place, sorting, tidying, precision handling, and multi-step manipulation across household, kitchen, office, and laboratory environments.
Available data modalities depend on the selected package and may include RGB video, 6-DoF gripper trajectories, gripper action/state streams, timestamps, and task-level descriptions. Structured episode delivery and LeRobot compatibility can be provided where validated.
Dataset card
# UMI Handheld Gripper — 100 Hours Robotic Manipulation Dataset
## Summary
Real-world robotic manipulation demonstrations collected using human-operated Universal Manipulation Interface (UMI) grippers. Designed for Vision-Language-Action (VLA) models, imitation learning, robotic policy training, and Embodied AI.
Dataset available for **on-demand collection**, with customizable tasks, environments, object diversity, and delivery volumes.
## Contents
- **Dataset category:** Robotic Manipulation
- **Collection type:** UMI handheld gripper
- **Total duration:** 100 hours (on-demand)
- **Episode count:** Approximately 3,000 (based on 120-second average)
- **Episode duration:** Greater than 90 seconds
- **Video resolution:** 1920 × 1080 minimum
- **Frame rate:** 30–60 FPS
- **Video codec:** H.264 / H.265, depending on capture configuration
- **Video format:** MP4
- **Total dataset size:** On demand, based on recording configuration
- **Audio:** On demand
- **Motion tracking:** 6-DoF gripper pose trajectories
- **Gripper actions:** Opening, closing, and manipulation states
- **Synchronization:** Within one video frame (target)
- **Dataset format:** LeRobot 3.0 on demand
- **Licence:** Commercial Standard, non-exclusive
## Manipulation Tasks
- Pick-and-place
- Grasping and object repositioning
- Lab scenarios
- Precision object handling
- Multi-step manipulation
- Household and laboratory activities
## Capture
Human operators perform manipulation tasks using handheld UMI grippers across real-world environments. Collection supports diverse objects, operators, manipulation scenarios, and task-level annotations.
## Layout
Episode-based RGB video recordings with associated 6-DoF gripper trajectories, action data, timestamps, and task metadata, according to the agreed collection specification.
## Intended Use
- Vision-Language-Action model training
- Robotic imitation learning
- Manipulation policy development
- Action prediction
- Sim-to-real robotics research
- Physical AI foundation model training
## Availability
**On-demand collection — 100-hour target package.** Custom volumes, task coverage, recording configurations, and delivery formats can be arranged based on buyer requirements.
