Paper Overview
- Field: Computer Vision (CV)
- Authors: Yizhou Xu, Lars Bretzner, Tiesheng Wang, Atsuto Maki
- Published: 2026-08-11
- arXiv: 2608.11203
- Proposes a self-supervised representation learning framework for understanding 3D skeleton-based human motion in soccer, with future motion prediction as the learning objective.
- Human motion is inherently uncertain, so accounting for multiple plausible futures is essential for capturing underlying motion dynamics and learning effective representations.
- Introduces a conditioning module for motion prediction that models a probabilistic distribution over discretized future motions in 3D Euclidean space, learning multimodality with explicit supervision from future trajectories.
- Experiments on large-scale soccer player tracking data show the approach substantially improves motion prediction accuracy.
- The learned representations transfer effectively to multiple soccer downstream applications, demonstrating strong cross-task generalization.
Key Points
Abstract (Original)
This paper presents a self-supervised representation learning framework for understanding 3D skeleton-based human motion in soccer, using future motion prediction as the learning objective. Since human motion is inherently uncertain, accounting for multiple plausible futures is essential for capturing the underlying motion dynamics and learning effective representations. To this end, we introduce a conditioning module for motion prediction that models a probabilistic distribution over discretized future motions in 3D Euclidean space, learning multimodality with explicit supervision from future trajectories. Experiments on large-scale soccer player tracking data show that our approach substantially improves motion prediction accuracy. Moreover, the learned representations effectively transfer to multiple downstream soccer applications.
*Paper link*: https://arxiv.org/abs/2608.11203