In this paper, we present a new bottom-up one-stage method for whole-body\npose estimation, which we call "hierarchical point regression," or HPRNet for\nshort. In standard body pose estimation, the locations of $\\sim 17$ major\njoints on the human body are estimated. Differently, in whole-body pose\nestimation, the locations of fine-grained keypoints (68 on face, 21 on each\nhand and 3 on each foot) are estimated as well, which creates a scale variance\nproblem that needs to be addressed. To handle the scale variance among\ndifferent body parts, we build a hierarchical point representation of body\nparts and jointly regress them. The relative locations of fine-grained\nkeypoints in each part (e.g. face) are regressed in reference to the center of\nthat part, whose location itself is estimated relative to the person center. In\naddition, unlike the existing two-stage methods, our method predicts whole-body\npose in a constant time independent of the number of people in an image. On the\nCOCO WholeBody dataset, HPRNet significantly outperforms all previous bottom-up\nmethods on the keypoint detection of all whole-body parts (i.e. body, foot,\nface and hand); it also achieves state-of-the-art results on face (75.4 AP) and\nhand (50.4 AP) keypoint detection. Code and models are available at\n\\url{https://github.com/nerminsamet/HPRNet}.\n