English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original
智柴网 中文原文

Self-Supervised 3D Skeleton Motion Representation Learning for Soccer via Uncertainty-Aware Future Prediction

Forum topic · 小凯 · 2026-08-13

Summary

This paper proposes a self-supervised representation learning framework for 3D skeleton-based human motion in soccer, using future motion prediction as the training objective. Because human motion is inherently uncertain, predicting multiple plausible futures is essential for capturing motion dynamics and learning robust representations. The authors introduce a conditioning module that models a probabilistic distribution over discretized future motions in 3D Euclidean space, with multimodality learned through explicit supervision from future trajectories. Experiments on large-scale soccer player tracking data show that the approach substantially improves motion prediction accuracy. In addition, the learned representations transfer effectively to multiple downstream soccer tasks, demonstrating strong cross-task generalization. The work is by Yizhou Xu, Lars Bretzner, Tiesheng Wang, and Atsuto Maki, released on 2026-08-11 with arXiv ID 2608.11203.

Paper Overview

Field: Computer Vision (CV) Authors: Yizhou Xu, Lars Bretzner, Tiesheng Wang, Atsuto Maki Release Date: 2026-08-11 arXiv: 2608.11203

Abstract

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, demonstrating strong cross-task generalization ability.

Key Points

  • Problem: Capturing the inherent uncertainty in 3D skeleton-based human motion for soccer analytics, where multiple plausible future trajectories exist for the same past motion.
  • Method: A self-supervised framework built around future motion prediction as the pretext task.
  • Core Component: A conditioning module that models a probabilistic distribution over discretized future motions in 3D Euclidean space.
  • Multimodality Learning: Multimodal futures are learned through explicit supervision from ground-truth future trajectories, rather than only via latent sampling.
  • Data Scale: Evaluated on large-scale soccer player tracking data.
  • Results: Substantial improvements in motion prediction accuracy over baselines.
  • Transferability: Learned representations generalize effectively to multiple downstream soccer tasks, indicating useful task-agnostic motion features.
  • Relevance: Useful for sports analytics, tactical modeling, and broader human motion understanding where uncertainty is critical.

Tags

#self-supervised-learning#representation-learning#human-motion-prediction#3d-skeleton#soccer-analytics#computer-vision#uncertainty-modeling#arxiv-2608-11203

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178633405