HRDexDB: A Large-Scale Dataset of Dexterous Human and Robotic Hand Grasps

2026-04-16 · Source: Computer Vision and Pattern Recognition · Field: Technology & Digital — Robotics & Autonomous Systems, Artificial Intelligence & Machine Learning, Computer Vision · Depth: Expert, quick

Summary

HRDexDB is a new large-scale, multi-modal dataset designed for dexterous grasping research, featuring both human and diverse robotic hands. It includes grasping trajectories for 100 distinct objects, captured using advanced vision techniques and a specialized multi-camera system. The dataset provides high-precision spatiotemporal 3D ground-truth motion for both the grasping agent and the manipulated object. HRDexDB integrates high-resolution tactile signals, synchronized multi-view video, and egocentric video streams to support physical interaction studies. Comprising 1.4K grasping trials, including successes and failures, the dataset offers visual, kinematic, and tactile modalities. It aims to serve as a foundational benchmark for multi-modal policy learning and cross-domain dexterous manipulation by aligning human and robotic dexterity on identical objects.

Key takeaway

For research scientists developing dexterous manipulation policies, HRDexDB offers a unique resource to train and evaluate models. You should explore its multi-modal data, including human and robotic grasps, to advance policy learning and benchmark cross-domain transfer. This dataset's comprehensive nature, including success and failure trials, can significantly improve the robustness of your robotic systems.

Key insights

HRDexDB is a multi-modal dataset for dexterous grasping, combining human and robotic hand data.

Principles

• Multi-modal data enhances dexterous manipulation research.
• Aligning human and robot data improves cross-domain learning.

Method

HRDexDB captures 1.4K grasping trials using a multi-camera system, generating 3D motion, tactile, and video data for human and robotic hands on 100 objects.

In practice

• Use HRDexDB for multi-modal policy learning.
• Benchmark cross-domain dexterous manipulation.

Topics

• HRDexDB
• Dexterous Grasping
• Robotic Manipulation
• Multi-modal Data
• Human-Robot Grasping

Best for: Research Scientist, AI Scientist, Robotics Engineer, Machine Learning Engineer

Related on AIssential

• Human Universal Grasping — Human egocentric grasping data and flow-matching models enable robots to achieve universal, zero-shot object manipulation.
• Do as I Do: Dexterous Manipulation Data from Everyday Human Videos — DO AS I DO reconstructs and retargets human video interactions to generate scalable, executable data for dexterous robotic manipulation.
• Is Diversity All You Need for Scalable Robotic Manipulation? — Not all data diversity benefits robotic manipulation equally; task diversity and debiased expert data are key.
• EquiDexFlow: Contact-Grounded SE(3)-Equivariant Dexterous Grasp Generative Flows — Jointly predicting grasp kinematics and contact forces ensures physically stable dexterous grasps.
• Video Friday: Humanoid Learns Tennis Skills Playing Humans — Robotics research is advancing dexterous manipulation, sustainable design, and complex locomotion through AI and novel hardware.
• Mana: Dexterous Manipulation of Articulated Tools — Mana reinterprets dexterous articulated tool manipulation as an animation problem, enabling zero-shot sim-to-real transfer via a coarse-to-fine pipeline.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computer Vision and Pattern Recognition.