Event
Galbot contributes DexGraspNet 2.0 large-scale dexterous grasping benchmark
Key points
- DexGraspNet 2.0 presents a large-scale synthetic benchmark for dexterous grasping in cluttered scenes, spanning 1,319 objects, 8,270 scenes, and 427 million grasps.
- The project page lists Galbot as an author affiliation and describes zero-shot sim-to-real transfer with 90.7 percent real-world dexterous grasping success in cluttered scenes.
- It matters because it is a major dataset and benchmark contribution in Galbot's dexterous-manipulation research lineage.
Company context
Develops embodied AI robotics systems focused on general-purpose manipulation and real-world deployment. The company builds integrated hardware platforms and vision-language-action models to enable robots to operate autonomously across commercial and industrial environments.
Context
- Company
- Galbot
- Segment
- Humanoid
- Event type
- Research Publication
- Geography
- Beijing · China