建模空间 - 时间交互以进行组合动作识别（英文版）

英文标题：Modelling Spatio-Temporal Interactions for Compositional Action Recognition中文摘要：该研究提出组合动作识别的方法，并使用时空交互编码器捕捉人体 - 物体之间的交互作用并将其与视频信息相结合。该方法在多个数据集上获得了先进的性能，表明显式地对人体 - 物体之间的交互行为建模是有效的。英文摘要：Humans have the natural ability to recognize actions even if the objectsinvolved in the action or the background are changed. Humans can abstract awaythe action from the appearance of the objects and their context which isreferred to as compositionality of actions. Compositional action recognitiondeals with i