This instruction will help you to pre-process the Human3.6M dataset - View it on GitHub