DexYCB is a dataset of human hand grasping objects, which can be used for tasks such as 2D target and key point detection, 6D target pose estimation, and 3D hand pose estimation. The data was acquired by 8 RGB-D cameras, recording RGB images and depth maps from 8 perspectives at 30fps with a resolution of 640*480. A total of more than 1,000 sequences were recorded, including 582,000 RGB-D frames.