· Research & technology· 日本語で読む
OmniDex expands dexterous grasping dataset to cluttered scenes
A research paper called “OmniDex,” which extends dexterous grasping to cluttered scenes, was posted on arXiv on Oct. 8.
What happened
The paper’s authors proposed a framework that organizes high-quality 3D objects and support surfaces and uses a progressive narrowing method, without relying on per-scene optimization. The paper says they built a benchmark containing more than 2.6 million scenes and 400 million ground-truth grasp data points.
Sources
Translated from the Japanese original: /news/omnidex-260-man-ken-chou-no-zatsuzen-kankyou-deno-haji-deeta-1c157b. The Japanese version is authoritative if they differ.