Bidirectional Temporal Diffusion Model for Temporally Consistent Human Animation

TL;DR

This paper presents a bidirectional temporal diffusion model that enhances the realism and temporal consistency of human animation generated from various inputs by suppressing motion ambiguity through bidirectional modeling.

Contribution

The paper introduces a novel bidirectional temporal diffusion framework for human animation, improving over unidirectional methods in temporal coherence and realism.

Findings

Outperforms existing unidirectional methods in temporal coherence

Generates realistic human animations from single images, videos, or noise

Effectively suppresses motion artifacts and appearance distortions

Abstract

We introduce a method to generate temporally coherent human animation from a single image, a video, or a random noise. This problem has been formulated as modeling of an auto-regressive generation, i.e., to regress past frames to decode future frames. However, such unidirectional generation is highly prone to motion drifting over time, generating unrealistic human animation with significant artifacts such as appearance distortion. We claim that bidirectional temporal modeling enforces temporal coherence on a generative network by largely suppressing the motion ambiguity of human appearance. To prove our claim, we design a novel human animation framework using a denoising diffusion model: a neural network learns to generate the image of a person by denoising temporal Gaussian noises whose intermediate results are cross-conditioned bidirectionally between consecutive frames. In the experiments, our method demonstrates strong performance compared to existing unidirectional approaches with realistic temporal coherence.