Skeleton-Based Action Recognition Based on Distance Vector and Multihigh View Adaptive Networks

Figure 3

MHNet schematic diagram. (a–c) New perspectives at different heights. Among them, (a) is the best view determined by the adaptive network, and (b) and (c) are the best view under the condition of adding angle constraints. From frame 1 to frame t the posture of the human body is constantly changing, and the MHNet can always find the best view for learning.