DexYCB

DexYCB is a benchmark dataset for capturing hand grasping of objects, introduced in 2021 by researchers from NVIDIA and University of Washington. It provides synchronized multi-view video of human hands grasping objects from the YCB object set, with comprehensive 3D annotations. The dataset supports three evaluation tasks: 2D object and keypoint detection, 6D object pose estimation, and 3D hand pose estimation. It also introduces a unique robotics-relevant task: generating safe robot grasps for human-to-robot object handover. DexYCB is widely used in dexterous manipulation research, hand-object interaction understanding, and human-robot collaboration. It bridges the gap between computer vision research on hand pose estimation and practical robotics applications requiring safe, stable object handover between humans and robots.