Fig. 5.
Visualization results of predictions for the four actions of (a) walking (b) smoking (c) walkingdog (d) greeting (e) eating (f) phoning on Human3.6 M. The ground-truth, LTD [4], HRI [27], and the proposed method are shown from top to bottom. The changes in actions from the first to the last frame of the prediction can be clearly seen in the grey dashed box, while the blue round box shows the comparison between predicted action and ground truth by the proposed method and other methods. It can be seen from the visualization results that the proposed method produces predictions closer to the ground truth than HRI and LTD.