Synthetic Object Placement Dataset
收藏arXiv2025-09-30 收录
下载链接：
https://any-place.github.io
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是一组综合性的数据，涵盖了机器人操作任务中的各种局部放置布局，包括堆叠、悬挂和插入等配置。该数据集利用Blender进行3D物体创建，并使用NVIDIA IsaacSim生成姿态，确保了放置场景的多样化表现。数据集规模涉及1489个物体，分为13个类别，共包含5370个放置姿态。其任务是针对机器人操作的对象放置预测。
This is a comprehensive dataset covering various local placement layouts in robotic manipulation tasks, including configurations such as stacking, hanging, and insertion. 3D objects are created using Blender, and poses are generated with NVIDIA IsaacSim to ensure diverse representations of placement scenarios. The dataset includes 1489 objects categorized into 13 classes, with a total of 5370 placement poses. The core task of this dataset is object placement prediction for robotic manipulation.
