DexGYS / Grasp as You Say

DOF: active and total degrees of freedom are distinct; neither alone measures dexterity.
Mass: read the single-hand boundary and whether external actuation, cables or controllers are included.
Price: check currency, date, quantity, taxes, controller and service inclusions. A public observation is not a current quotation.
Performance: compare the same task, conditions and success definition. Different demonstrations do not form a fair leaderboard.

iSEE Laboratory

Dexterous grasp labels paired with language guidance for learning instruction-conditioned grasp generation.

What you can explore

Explore the dataset alongside the authors’ grasp-generation code and evaluation setup.

The workflow needs separate Shadow Hand models and OakInk object meshes. A language-matched grasp still requires collision and physical validation.

Yi-Lin Wei, Jian-Jian Jiang, Chengyi Xing, Xiantuo Tan, Xiao-Ming Wu, Hao Li, Mark Cutkosky and Wei-Shi Zheng

License / access: Dataset card: MIT · source object meshes and hand assets have separate terms

Editorial reference · checked 2026-09-22