HUG: dexterous grasping trained on zero robot data, already running on robots worldwide
LerrelPinto · x · 2026-09-05
Kevin Wu's Human Universal Grasping (HUG) was rejected from CoRL 2026, yet Lerrel Pinto notes it's already deployed globally — including on LG's CLOiD robot in Seoul and Meta Aria in Seattle.
Key points:
- Trained solely on real-world human hand grasping data with zero robot data, HUG generates diverse human-hand grasps for any object in an RGB-D image from a stereo camera
- Works out of the box with any stereo camera, generalizing to unseen objects and environments
- Fully open-sourced: paper, inference/visualization and training code, the 1M-HUGs dataset, and HUG-Bench; supports RealSense and ZeD live streaming on Ubuntu 22.04/24.04, CUDA 12.8, PyTorch 2.9.1
More from Embodied
- New tendon-based robotic hand designs trained with PPO to learn one-armed catch — _krishna_murthy · 2026-10-23
- Rivian acquires world model startup Third Dimension to accelerate its self-driving push — emmanuelvivier · 2026-10-23
- Nvidia's Berlin GTC spotlights humanoid robots and Vera Rubin AI supercomputing — emmanuelvivier · 2026-10-23
- XPeng commissions 80%-automated line for IRON humanoid, mass production by end-2026 — ingliguori · 2026-10-23
- HKU's TILDE Distills Embodied Foundation Models into Robot Policies — hkuhk · 2026-10-23
- 10 of 12 Beta Robots Worked Without Fine-Tuning — Do We Still Need It? — micoolcho · 2026-10-23