Published July 13, 2024 | Version v1
Journal article Open

TEDi: Temporally-Entangled Diffusion for Long-Term Motion Synthesis

  • 1. University of Chicago
  • 2. Snap Inc.

Description

The gradual nature of a diffusion process that synthesizes samples in small increments constitutes a key ingredient of Denoising Diffusion Probabilistic Models (DDPM), which have presented unprecedented quality in image synthesis and been recently explored in the motion domain. In this work, we propose to adapt the gradual diffusion concept (operating along a diffusion time-axis) into the temporal-axis of the motion sequence. Our key idea is to extend the DDPM framework to support temporally varying denoising, thereby entangling the two axes. Using our special formulation, we iteratively denoise a motion buffer that contains a set of increasingly-noised poses, which auto-regressively produces an arbitrarily long stream of frames. With a stationary diffusion time-axis, in each diffusion step we increment only the temporal-axis of the motion such that the framework produces a new, clean frame which is removed from the beginning of the buffer, followed by a newly drawn noise vector that is appended to it. This new mechanism paves the way towards a new framework for long-term motion synthesis with applications to character animation and other domains.

Files

TEDi.pdf

Files (239.4 MB)

Name Size Download all
md5:912d7ab6404f1a30fe8ec1bcb75fe66d
214.5 MB Preview Download
Article
md5:da75cf69ff5cedf4c7bd90df3a4cf5ad
24.9 MB Preview Download

Additional details

Identifiers

DOI
10.1145/3641519.3657515
Other
oai:uchicago.tind.io:12817

Funding

National Science Foundation
2241303

UChicago Information

Division(s)
Physical Sciences Division
Department(s)
Computer Science