Dyadic Human Motion Prediction
This addresses motion prediction for interacting humans, such as in dance scenarios, and is incremental as it extends existing methods to dyadic interactions.
The paper tackles the problem of predicting human motion for closely interacting subjects, which prior single-person methods fail to account for, by introducing a framework with a pairwise attention mechanism and a new dance dataset, resulting in outperforming state-of-the-art single-person techniques.
Prior work on human motion forecasting has mostly focused on predicting the future motion of single subjects in isolation from their past pose sequence. In the presence of closely interacting people, however, this strategy fails to account for the dependencies between the different subject's motions. In this paper, we therefore introduce a motion prediction framework that explicitly reasons about the interactions of two observed subjects. Specifically, we achieve this by introducing a pairwise attention mechanism that models the mutual dependencies in the motion history of the two subjects. This allows us to preserve the long-term motion dynamics in a more realistic way and more robustly predict unusual and fast-paced movements, such as the ones occurring in a dance scenario. To evaluate this, and because no existing motion prediction datasets depict two closely-interacting subjects, we introduce the LindyHop600K dance dataset. Our results evidence that our approach outperforms the state-of-the-art single person motion prediction techniques.