Lie-X: Depth Image Based Articulated Object Pose Estimation, Tracking, and Action Recognition on Lie Groups

摘要

Pose estimation, tracking, and action recognition of articulated objects fromdepth images are important and challenging problems, which are normallyconsidered separately. In this paper, a unified paradigm based on Lie grouptheory is proposed, which enables us to collectively address these relatedproblems. Our approach is also applicable to a wide range of articulatedobjects. Empirically it is evaluated on lab animals including mouse and fish,as well as on human hand. On these applications, it is shown to delivercompetitive results compared to the state-of-the-arts, and non-trivialbaselines including convolutional neural networks and regression forestmethods.