Dyadic Human Motion Prediction

TL;DR

This paper presents a novel human motion prediction framework that explicitly models interactions between two subjects using a pairwise attention mechanism, improving predictions in scenarios like dance.

Contribution

The paper introduces a new framework for dyadic human motion prediction that accounts for mutual dependencies, along with a new dataset for two-person interactions.

Findings

Outperforms state-of-the-art single-person motion prediction methods.

Introduces LindyHop600K dataset for two-person dance interactions.

Enhances prediction of fast-paced and unusual movements.

Abstract

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.