Summary
FlowMoP is a conditional flow matching framework for stochastic multi-person 3D motion prediction, presented in an arXiv paper by Aadya Agrawal, Ho Kei Cheng, and Alexander Schwing. Unlike the dominant deterministic paradigm that outputs a fixed number of predictions per agent, FlowMoP jointly generates diverse future motions for multiple agents, modeling both global trajectories and root-relative poses. A B-spline reparameterization compresses the generative state space while enforcing temporal smoothness, and a dual-stream transformer conditions each agent's predictions on observed neighbor motion, enabling interaction-aware forecasting. The method achieves state-of-the-art results on CMU-Mocap (UMPM) and 3DPW, surpassing deterministic baselines on joint position error, pose error, and final displacement error—while producing full predictive distributions. The authors also establish the first motion prediction baseline on WorldPose, a large-scale professional soccer dataset, where social conditioning notably concentrates predictive uncertainty in interaction-constrained directions.
Overview
Field: Computer Vision
Authors: Aadya Agrawal, Ho Kei Cheng, Alexander Schwing
Published: 2026-11-06
arXiv: 2611.09882
Summary
Predicting the future motion of multiple people is inherently stochastic: the same observed history admits many plausible continuations depending on intent, coordination, and social context. Yet the dominant paradigm in multi-person 3D motion prediction remains deterministic, with existing methods producing a fixed number of outputs per agent.
FlowMoP addresses this with a conditional flow matching framework that jointly generates diverse future motions for multiple agents, modeling both global trajectories and root-relative poses. Key design elements:
- B-spline reparameterization: compresses the generative state space while enforcing temporal smoothness.
- Dual-stream transformer: conditions each agent's predictions on observed neighbor motion for interaction-aware forecasting.
Results
- State-of-the-art performance on CMU-Mocap (UMPM) and 3DPW, outperforming deterministic baselines on joint position error, pose error, and final displacement error—while producing complete predictive distributions.
- First motion prediction baseline established on WorldPose, a large-scale professional soccer dataset, where social conditioning clearly concentrates predictive uncertainty along interaction-constrained directions.
---
*Auto-collected on 2026-11-10*
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/178636978