HumanCM: One Step Human Motion Prediction

2025/10/19 by Haojie Liu, Haojie, Liu, Suixiang, Gao
Computer Science · Engineering · Health Professions · #Human Pose and Action Recognition #Human Motion and Animation #Balance, Gait, and Falls Prevention

paper · pdf · doi:10.48550/arxiv.2510.16709

Abstract

We present HumanCM, a one-step human motion prediction framework built upon consistency models. Instead of relying on multi-step denoising as in diffusion-based methods, HumanCM performs efficient single-step generation by learning a self-consistent mapping between noisy and clean motion states. The framework adopts a Transformer-based spatiotemporal architecture with temporal embeddings to model long-range dependencies and preserve motion coherence. Experiments on Human3.6M and HumanEva-I demonstrate that HumanCM achieves comparable or superior accuracy to state-of-the-art diffusion models while reducing inference steps by up to two orders of magnitude.

Citations

• Simplifying, Stabilizing and Scaling Continuous-Time Consistency Models
• MotionLCM: Real-time Controllable Motion Generation via Latent Consistency Model
• Latent Consistency Models: Synthesizing High-Resolution Images with Few-Step Inference
• TransFusion: A Practical and Effective Transformer-based Diffusion Model for 3D Human Motion Prediction
• Consistency Models
• HumanMAC: Masked Motion Completion for Human Motion Prediction
• Human Joint Kinematics Diffusion-Refinement for Stochastic Motion Prediction
• Classifier-Free Diffusion Guidance
• Diverse Human Motion Prediction via Gumbel-Softmax Sampling from an Auxiliary Space
• DPM-Solver: A Fast ODE Solver for Diffusion Probabilistic Model Sampling in Around 10 Steps
• Space-Time-Separable Graph Convolutional Network for Pose Forecasting
• Generating Smooth Pose Sequences for Diverse Human Motion Prediction
• We are More than Our Joints: Predicting how 3D Bodies Move
• Score-Based Generative Modeling through Stochastic Differential Equations
• Denoising Diffusion Implicit Models
• Denoising Diffusion Probabilistic Models
• DLow: Diversifying Latent Flows for Diverse Human Motion Prediction
• Learning Trajectory Dependencies for Human Motion Prediction
• Diverse Trajectory Forecasting with Determinantal Point Processes
• Context-aware Human Motion Prediction
• Uncertainty-aware Short-term Motion Prediction of Traffic Actors for\n Autonomous Driving
• MT-VAE: Learning Motion Transformations to Generate Multimodal Human Dynamics
• Accurate and Diverse Sampling of Sequences based on a "Best of Many" Sample Objective
• HP-GAN: Probabilistic 3D human motion prediction via GAN
• Auto-Conditioned Recurrent Networks for Extended Complex Human Motion Synthesis
• DeLiGAN : Generative Adversarial Networks for Diverse and Limited Data
• On human motion prediction using recurrent neural networks
• A Survey of Motion Planning and Control Techniques for Self-driving\n Urban Vehicles
• Recurrent Network Models for Human Dynamics

Related

• HumanMAC: Masked Motion Completion for Human Motion Prediction
• DivDiff: A Conditional Diffusion Model for Diverse Human Motion Prediction
• Towards Accurate Human Motion Prediction via Iterative Refinement
• MotionPCM: Real-Time Motion Synthesis with Phased Consistency Model
• CoMusion: Towards Consistent Stochastic Human Motion Prediction via Motion Diffusion