Robotics

CosmoH2G Turns Flips and Rotations Into Gripper Motions

Shar Hendrix 4 min read

Most “copy this human video” pipelines choke when the object rotates or flips. CosmoH2G, posted 7 September on arXiv by Hongxiang Zhao, Mutian Xu, and colleagues at CUHKSZ, FNii-Shenzhen, and GenuX, is a paired hand-to-gripper dataset built for those motions.

6,189 episodes. 1,254 objects. Handheld UMI grippers mimicking the same contact patch, orientation, and path as the human hand.

Five-frame human demonstration of rotating a bottle compared with matching robotic gripper execution
Human demo on top, gripper replay below, including a rotation in clutter. Source: Zhao, Xu et al., arXiv:2609.07498, Figure 1.

Why a new dataset

Rule-based retargeting (thumb–index midpoint as the grasp) fails on natural grasps. Object-flow methods lose tracked points when a face rotates out of view. The authors’ protocol forces in-plane rotations and vertical flips, varies grasp type and contact area, and keeps start/end object poses aligned between hand and UMI takes. Pairs below 0.9 trajectory similarity are dropped; retained pairs average 0.957.

Object mix from their Table 1: toys 33.33%, decor 27.88%, food 23.03%, daily necessities 12.12%, beauty 3.64%. Code and dataset are listed as coming soon on the project page.

Paired RGB-D hand and UMI gripper videos with 3D reconstruction pipeline
Paired hand and UMI captures, then 3D registration. Source: arXiv:2609.07498, Figure 5.

Two stages, then a cleanup

End-to-end pose sequences drifted. Their fix:

  1. Stage I predicts only start and terminal 6-DoF gripper poses from hand and object points.
  2. Stage II learns orientation from the paired data. Translation is initialized from the hand path (aligned contact regions) and then optimized for contact, smoothness, and inverse-kinematics feasibility.

Test split: 186 cases on 40 unseen objects. Simulation is GalaxeaManipSim on an R1 Lite. Hardware is the same wheeled dual-arm.

MethodSim successReal success
MimicFunc56.45%43.01%
3DFlowAction44.62%36.02%
Im2Flow2Act57.53%51.61%
Track2Act (retrained)77.42%60.22%
CosmoH2G83.87%70.43%

Real-robot drops: objects slip during large rotations, or get dragged before they leave the table.

Real-robot comparison: CosmoH2G places a bottle upright while four baselines miss orientation or position
Galaxea R1 Lite: CosmoH2G vs four baselines on a rotation-and-place task. Source: arXiv:2609.07498, Figure 7.

Limits they list: open-loop, no live collision avoidance, pick-and-place only. A one-stage model is left for when the paired set is larger.

A Human’s Take

The useful part is the collection protocol, not the transformer. If you force flips in the data, the mapping has something to learn. Seventy percent on a real R1 Lite with unseen objects is enough to keep watching. Closed-loop is the next receipt I want.

Sources