Capturing Uncertainty in Human Motion for Representation Learning in Soccer (arXiv 2608.11203)

Forum topic · 小凯 · 2026-08-13

Summary

This paper introduces a self-supervised representation learning framework for understanding 3D skeleton-based human motion in soccer, using future motion prediction as the learning objective. Because human motion is inherently uncertain, the authors argue that accounting for multiple plausible futures is essential for capturing underlying motion dynamics and learning effective representations. They propose 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, and the learned representations transfer effectively to multiple soccer-related downstream applications, demonstrating strong cross-task generalization. Authored by Yizhou Xu, Lars Bretzner, Tiesheng Wang, and Atsuto Maki; available on arXiv as 2608.11203.

Paper Overview

Field: Computer Vision Authors: Yizhou Xu, Lars Bretzner, Tiesheng Wang, Atsuto Maki Posted: 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, the authors 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.

Key Findings

Experiments on large-scale soccer player tracking data show that the approach substantially improves motion prediction accuracy.
The learned representations effectively transfer to multiple soccer-related downstream applications.
The results demonstrate strong cross-task generalization of the learned representations.

--- *Auto-collected on 2026-08-13.*