Learning to Predict Diverse Human Motions from a Single Image via Mixture Density Networks

TL;DR

This paper introduces a novel method using mixture density networks to predict diverse future human motions from a single image, addressing the challenge of limited input information and stochastic motion uncertainty.

Contribution

The approach enables generation of multiple plausible human motions from a single image, incorporating energy-based loss functions for improved coherence and accuracy.

Findings

Effective in generating diverse motion hypotheses

Achieves high prediction accuracy on benchmark datasets

Outperforms existing methods in diversity and coherence

Abstract

Human motion prediction, which plays a key role in computer vision, generally requires a past motion sequence as input. However, in real applications, a complete and correct past motion sequence can be too expensive to achieve. In this paper, we propose a novel approach to predicting future human motions from a much weaker condition, i.e., a single image, with mixture density networks (MDN) modeling. Contrary to most existing deep human motion prediction approaches, the multimodal nature of MDN enables the generation of diverse future motion hypotheses, which well compensates for the strong stochastic ambiguity aggregated by the single input and human motion uncertainty. In designing the loss function, we further introduce the energy-based formulation to flexibly impose prior losses over the learnable parameters of MDN to maintain motion coherence as well as improve the prediction accuracy by customizing the energy functions. Our trained model directly takes an image as input and generates multiple plausible motions that satisfy the given condition. Extensive experiments on two standard benchmark datasets demonstrate the effectiveness of our method in terms of prediction diversity and accuracy.