DexGraspNet: A Large-Scale Robotic Dexterous Grasp Dataset for General Objects Based on Simulation

TL;DR

DexGraspNet addresses the lack of large-scale dexterous grasp datasets by generating 1.32 million ShadowHand grasps on 5355 objects across 133+ categories using a deeply accelerated, differentiable force-closure pipeline. The method introduces initialization strategies, a reverse penetration energy, and penalties to yield diverse, high-quality grasps, validated in Isaac Gym and applicable to multiple dexterous hands. Cross-dataset experiments show that training on DexGraspNet improves grasp quality and diversity over the prior DDG data, confirming its value for advancing dexterous grasping research. The dataset and code are released to empower future work in dexterous manipulation and generalization across hands like ShadowHand, MANO, and Allegro.

Abstract

Robotic dexterous grasping is the first step to enable human-like dexterous object manipulation and thus a crucial robotic technology. However, dexterous grasping is much more under-explored than object grasping with parallel grippers, partially due to the lack of a large-scale dataset. In this work, we present a large-scale robotic dexterous grasp dataset, DexGraspNet, generated by our proposed highly efficient synthesis method that can be generally applied to any dexterous hand. Our method leverages a deeply accelerated differentiable force closure estimator and thus can efficiently and robustly synthesize stable and diverse grasps on a large scale. We choose ShadowHand and generate 1.32 million grasps for 5355 objects, covering more than 133 object categories and containing more than 200 diverse grasps for each object instance, with all grasps having been validated by the Isaac Gym simulator. Compared to the previous dataset from Liu et al. generated by GraspIt!, our dataset has not only more objects and grasps, but also higher diversity and quality. Via performing cross-dataset experiments, we show that training several algorithms of dexterous grasp synthesis on our dataset significantly outperforms training on the previous one. To access our data and code, including code for human and Allegro grasp synthesis, please visit our project page: https://pku-epic.github.io/DexGraspNet/.

Figures (6)

• Figure 1: A visualization of DexGraspNet. DexGraspNet contains 1.32M grasps of ShadowHandshadowhand on 5355 objects, which is two orders of magnitudes larger than the previous dataset from DDG liu2020deep. It features diverse types of grasping that cannot be achieved using GraspIt!graspit.
• Figure 2: (a) Green spheres with the radius of $\delta=1{\rm cm}$ are manually selected to compute $E_{\rm spen}$. (b) Contact candidates on the collision mesh in the canonical initial hand pose. (c) Initialization: 1. sample points on the object's inflated convex hull (blue); 2. move the hands to the sampled points and jitter the translation, rotation, and joint angles.
• Figure 3: Distribution of object numbers with respect to the average success rate for each object after final validation. We only keep successful grasps in our dataset.
• Figure 4: Visualization of grasps using different dexterous hands. From left to right: ShadowHand, MANO, Allegro.
• Figure 5: Visualization of the diverse grasps on the objects from DexGraspNet.