Skeleton-Based Mutually Assisted Interacted Object Localization and Human Action Recognition

Skeleton data carries valuable motion information and is widely explored in human action recognition. However, not only the but also interaction with environment provides discriminative cues to recognize of persons. In this paper, we propose a joint learning framework for mutually assisted ''interacted object localization'' ''human recognition'' based on skeleton data. The two tasks are serialized together collaborate promote each other, where preliminary type derived from alone helps improve interacted localization, which turn final Besides, explore temporal consistency as constraint better localize absence ground-truth labels. Extensive experiments datasets SYSU-3D, NTU60 RGB+D, Northwestern-UCLA UAV-Human show that our method achieves best or competitive performance state-of-the-art methods Visualization results can provide reasonable localization results.