airvlab/Grasp-Anything-6D

Grasp-Anything-6D数据集是一个大规模的数据集，专为语言驱动的6自由度抓取检测任务设计。该数据集基于之前的Grasp-Anything数据集，包含了点云场景、3D掩码、6自由度抓取姿势和抓取指令。点云场景文件为.npy格式，每个场景包含8192个点；3D掩码文件也是.npy格式；6自由度抓取姿势文件为.pkl格式，包含变换矩阵和抓取宽度；抓取指令文件同样为.pkl格式。

The Grasp-Anything-6D dataset is a large-scale dataset designed for the task of language-driven 6-DoF grasp detection. Building upon the previous Grasp-Anything dataset, it includes point cloud scenes, 3D masks, 6-DoF grasp poses, and grasp instructions. The point cloud scenes are stored in .npy files, each containing 8192 points; the 3D masks are also in .npy format; the 6-DoF grasp poses are in .pkl files, containing transformation matrices and grasp widths; and the grasp instructions are also in .pkl format.