The authors built the OmniDex benchmark by first curating high-quality 3D objects and supporting bases, then generating free-space seed grasps for single objects and filtering them through collision constraints in cluttered scenes using gravity-stabilized simulations.
Large-scale benchmark and end-to-end model achieve robust dexterous grasping in cluttered scenes
OmniDex generates 2.6 million cluttered scenes with 0.4 billion validated grasp poses, enabling a model with physics-constrained learning to outperform prior methods without post-optimization.
Top University
Naiyu Fang · Zhongjin Luo · Yuxin Mo · Siyuan Huang · Jianbo Liu · Yufei Liu · +4 more
ACE Robotics · CUHK, MMLab · CPII under InnoHK · Tongji University · Shanghai Jiao Tong University
Research Digest··2 min read
The authors introduce OmniDex, a massive simulation benchmark for dexterous grasping in cluttered scenes, created via a scalable seed-and-filter strategy that avoids costly scene-level optimization.
Why this paper
From CUHK, MMLab and 6 others
In one line
A 2.6-million-scene benchmark and end-to-end model achieve robust dexterous grasping in cluttered scenes.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§