Inspire Robotics, working with Shanghai Jiao Tong University and Fudan University, has open-sourced HandEdit, a dataset and benchmark for first-person human-to-robot dexterous-hand image editing.
What HandEdit actually is
The task: take a first-person human-operated video and replace the hand or arm with a specified robot configuration, while keeping the object, contact and background intact. That means fully removing the original hand and matching the target URDF’s structure, material and joints.
The dataset spans more than 200 million image-edit samples and 300,000-plus video clips, drawn from five egocentric datasets, EgoDex, ARCTIC, OakInk2, HOI4D and HO-Cap. It covers 26 URDF configurations, 13 standalone dexterous hands and 13 hand-arm forms, across 600-plus scenes, 1,100-plus objects and 400-plus task types. The 13 hands include Inspire’s own RH56DFX and RH5DG2, while the arm set pairs with JAKA, KUKA, Panda, RM65/75, UR5 and xArm platforms.
How the data is made
A pipeline segments the hand with SAM3, inpaints the background with ProPainter, redirects the pose to the target URDF via MANO or 3D handpose, solves arm inverse kinematics, renders under a matched first-person camera and screens out bad samples. Beyond synthetic images, it keeps structured joint states for checking configuration and correspondence.
Two evaluation tracks, Hand-only and Hand-Arm, each ship with 1,000 test images. For a field where dexterous-operation data is scarce and tied to specific robot bodies, a shared, open benchmark lowers the cost of comparing methods and should accelerate manipulation research well beyond any single lab.
Editor’s note: This is an adapted translation of the original OFweek Robotics report. It has been trimmed and restructured for readability for an international business audience. The full original (in Chinese) is at https://robot.ofweek.com/2026-09/ART-898890-8120-30701741.html.
Translated and adapted from OFweek Robotics (https://robot.ofweek.com/2026-09/ART-898890-8120-30701741.html).