AMASS, the full name of which is Archive of Motion Capture As Surface Shapes, is a large-scale human motion posture dataset.

The authors used the SMPL model to unify the parameters of multiple datasets (such as CMU, MPI-HDM05, MPIPose Limits, KIT, BioMotion Lab, TCD, and ACCAD), and finally integrated them into a new dataset: AMASS. This dataset is much richer than previous human motion datasets, including more than 40 hours of motion data, covering more than 300 subjects, and more than 11,000 actions.

AMASS can be used for animation, visualization, and generating training data for deep learning.