PartDrag-4D
收藏资源简介:
PartDrag-4D数据集是一个大规模的4D动态数据集,由清华大学、密歇根大学、北京大学和BAAI项目共同开发。该数据集包含了超过20,000个状态下的多视角观察数据,基于PartNet-Mobility数据集构建,提供静态物体的部件级运动的多视角图像。数据集覆盖了8个类别的738个网格,通过在不同位置渲染物体的运动状态来生成。它旨在为机器人操作任务提供支持,并推动部件级运动学习领域的研究。
PartDrag-4D Dataset is a large-scale 4D dynamic dataset jointly developed by Tsinghua University, University of Michigan, Peking University, and the BAAI project. This dataset contains over 20,000 sets of multi-view observation data across different states, is constructed based on the PartNet-Mobility dataset, and provides multi-view images of part-level motions for static objects. The dataset covers 738 meshes across 8 categories, and is generated by rendering the motion states of objects at various positions. It aims to support robotic manipulation tasks and promote research in the field of part-level motion learning.




