Something-Else: Compositional Action Recognition With Spatial-Temporal Interaction Networks

• Published in: Proceedings (IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Online) pp. 1046 - 1056
• Main Authors: Materzynska, Joanna, Xiao, Tete, Herzig, Roei, Xu, Huijuan, Wang, Xiaolong, Darrell, Trevor
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.06.2020
• Subjects: Cognition; Computational modeling; Detectors; Feature extraction; Task analysis; Training; Videos

Human action is naturally compositional: humans can easily recognize and perform actions with objects that are different from those used in training demonstrations. In this paper, we study the compositionality of action by looking into the dynamics of subject-object interactions. We propose a novel...