Unite The People is an image dataset for 3D human pose estimation. These images come from Leeds Sports Pose Datasetand its extended version, and MPII Human Pose DatasetEach image has different types of annotations, such as segmentation labels, poses, or 3D, and the human body model in the image is divided into 31 parts and 91 key points.

The data, code, and models involved in this dataset can be used to study human pose estimation, and the 3D model can be used to train a 2D human pose estimator.