Human Motion Modeling using DVGANs

TL;DR

This paper introduces DVGANs, a new GAN-based model for human motion generation and completion that is robust to noise, generalizes across actions, and produces diverse long sequences.

Contribution

The paper proposes a novel GAN architecture with dense validation and input perturbation for translation invariance, enhancing human motion modeling capabilities.

Findings

Resilient to noise in motion data

Generalizes well across different actions

Capable of generating long, diverse motion sequences

Abstract

We present a novel generative model for human motion modeling using Generative Adversarial Networks (GANs). We formulate the GAN discriminator using dense validation at each time-scale and perturb the discriminator input to make it translation invariant. Our model is capable of motion generation and completion. We show through our evaluations the resiliency to noise, generalization over actions, and generation of long diverse sequences. We evaluate our approach on Human 3.6M and CMU motion capture datasets using inception scores.