NVIDIA / University of Washington
DexYCB
Multi-view observations of people grasping objects, with hand and object pose benchmarks.
DEXTEROUS MANIPULATION
The world of dexterous hands.
Compare hardware. Discover applications. Share what works. An open reference and community for dexterous manipulation.
Datasets
Explore source-linked manipulation datasets on Dexhands.
PKU-EPIC
DexGraspNet
Synthetic dexterous grasp configurations generated for a Shadow Hand and validated in simulation.
Max Planck Institute for Intelligent Systems
GRAB
Whole-body human motion and object interactions, with articulated hand poses and derived contact.
Stanford research team
DexCap
Hand motion capture and scene observations for learning dexterous manipulation from human demonstrations.
DROID Dataset Team
DROID
Robot demonstrations collected in varied scenes using a shared Franka manipulation setup.
Open X-Embodiment collaboration
Open X-Embodiment
A collection of robot-learning datasets spanning multiple institutions and embodiments.
ETH Zürich / University of Amsterdam / MPI-IS
ARCTIC
Two human hands interacting with articulated objects, with motion and contact information for reconstruction research.
Shanghai Jiao Tong University / South China University of Technology
OakInk2
Bimanual human manipulation recordings with task structure, hand–object motion and object affordance annotations.
NVIDIA / DexMimicGen research team
DexMimicGen
Generated demonstrations and simulation environments for learning coordinated dexterous manipulation with two arms.
Meta research
HOT3D
Multi-view egocentric recordings of human hands interacting with objects, with tracking annotations.
Brown IVL
GigaHands
Bimanual human hand motion with text descriptions and visual observations.
Meta research / ContactPose team
ContactPose
Human grasps paired with hand and object pose, images and object contact observations.
ContactDB research team
ContactDB
Thermal contact observations showing how people grasp three-dimensional objects.
HO-3D research team
HO-3D
Images and three-dimensional annotations of a human hand manipulating objects under occlusion.
HOI4D research team
HOI4D
Egocentric human–object interactions with RGB-D observations, geometry, pose and segmentation annotations.
OakInk research team
OakInk
Hand–object interaction resources connecting object shape, grasp knowledge and visual observations.
H2O research team
H2O
Two-hand and object observations for egocentric interaction and action understanding.
TACO research team
TACO
Human tool–action–object interactions with coordinated hands and object motion.
SNU VCLab
ParaHome
Parameterized human activity in home environments, including body, hands and articulated objects.
RH20T research team
RH20T
Multimodal robot manipulation recordings with paired human demonstrations and sensing information.
4DVLab / RealDex team
RealDex
Real-world dexterous robot grasp recordings for learning human-like manipulation.
Meta research / InterHand team
InterHand2.6M
Images and annotations for three-dimensional pose estimation of single and interacting human hands.
AssemblyHands research team
AssemblyHands
Egocentric hand-pose resources for assembly activities and interaction understanding.
HoloAssist research team
HoloAssist
Human task demonstrations with instructor dialogue and multimodal observations for interactive assistance.
GenDexGrasp research team
MultiDex
Synthetic grasp configurations for multiple hand embodiments, used by GenDexGrasp.
Seoul National University / HRDexDB team
HRDexDB
Paired human and robot grasp recordings with visual, kinematic and three-dimensional annotations.
HandEdit research team
HandEdit
A dataset and benchmark for replacing visible human hands or arms with specified robot embodiments in egocentric images.
SoftVTBench research team
SoftVTBench
Visuo-tactile simulation demonstrations for rigid and deformable-object manipulation, with benchmark code and assets.
NeuralFeels research team
FeelSight
Vision, touch and proprioceptive observations recorded during in-hand object rotation.
iSEE Laboratory
DexGYS / Grasp as You Say
Dexterous grasp labels paired with language guidance for learning instruction-conditioned grasp generation.