COCO WholeBody is an extension of the COCO dataset that adds dense whole-body keypoint annotations, covering the body, hands, feet, and face with 133 keypoints per person rather than the 17 body-only keypoints of the original COCO annotations. It is used to train and benchmark whole-body pose estimation models, including hand and facial landmark detectors used by tools such as DWPose. The extended annotation set enables downstream applications such as gesture recognition, sign-language interpretation, and fine-grained human motion capture from images.