COCO-WholeBody

CC-BY-NC 4.0 ( not for commercial purpose)

COCO-WholeBody is an extension of COCO dataset with whole-body annotations. There are 4 types of bounding boxes (person box, face box, left-hand box, and right-hand box) and 133 keypoints (17 for body, 6 for feet, 68 for face and 42 for hands) annotations for each person in the image.

Source: Whole-Body Human Pose Estimation in the Wild

Image source: https://arxiv.org/pdf/2007.11858v1.pdf

Benchmarks

1 Image, 2*2 Stitchi/keypoint AP 2D Human Pose Estimation/WB 2D Human Pose Estimation/body 2D Human Pose Estimation/foot 2D Human Pose Estimation/face 2D Human Pose Estimation/hand 3D/keypoint AP 3D/AP 3D/AP50 3D/AP75 3D/APL 3D/APM 3D Face Modelling/AP 3D Face Modelling/AP50 3D Face Modelling/AP75 3D Face Modelling/APL 3D Face Modelling/APM 3D Face Modelling/keypoint AP 3D Face Reconstruction/AP 3D Face Reconstruction/AP50 3D Face Reconstruction/AP75 3D Face Reconstruction/APL 3D Face Reconstruction/APM 3D Face Reconstruction/keypoint AP Face Detection/AP Face Detection/AP50 Face Detection/AP75 Face Detection/APL Face Detection/APM Face Reconstruction/AP Face Reconstruction/AP50 Face Reconstruction/AP75 Face Reconstruction/APL Face Reconstruction/APM Face Reconstruction/keypoint AP Facial Landmark Detection/keypoint AP Facial Recognition and Modelling/AP Facial Recognition and Modelling/AP50 Facial Recognition and Modelling/AP75 Facial Recognition and Modelling/APL Facial Recognition and Modelling/APM Facial Recognition and Modelling/keypoint AP Hand/keypoint AP Hand Pose Estimation/keypoint AP Multi-Person Pose Estimation/keypoint AP Pose Estimation/keypoint AP