DexYCB
收藏资源简介:
DexYCB是由NVIDIA和华盛顿大学联合创建的数据集,专注于捕捉手部抓取物体的动作。该数据集包含582,000个RGB-D帧,覆盖1,000个序列,涉及10个参与者与20种不同物体的互动,从8个视角进行记录。数据集旨在支持3D物体姿态估计和3D手部姿态估计等任务的研究,特别适用于机器人学习和人机交互领域。通过多摄像头同步记录,结合人工标注,DexYCB提供了大规模的真实手部与物体交互数据,支持深度学习模型的训练和评估。
DexYCB is a dataset jointly created by NVIDIA and the University of Washington, focusing on capturing hand-object grasping motions. This dataset contains 582,000 RGB-D frames, covering 1,000 sequences involving interactions between 10 participants and 20 distinct objects, recorded from 8 camera viewpoints. It aims to support research on tasks such as 3D object pose estimation and 3D hand pose estimation, and is particularly suitable for the fields of robotics learning and human-robot interaction. Through synchronized multi-camera recording combined with manual annotation, DexYCB provides large-scale real-world hand-object interaction data to support the training and evaluation of deep learning models.

- 1HandoverSim: A Simulation Framework and Benchmark for Human-to-Robot Object Handovers华盛顿大学 · 2022年



