Fig. 3From: Flexible human action recognition in depth video sequences using masked joint trajectoriesExample image of the datasets used. Left: self-generated, center: CMU MoCap, right: MSR-Action3DBack to article page