We create D3D-HOI a dataset of monocular videos with ground truth annotations of 3D object pose and part motion during human-object interaction. - View it on GitHub