Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

Differentiable Dynamics for Articulated 3d Human Motion Reconstruction

Gärtner, Erik LU orcid ; Andriluka, Mykhaylo ; Coumans, Erwin and Sminchisescu, Cristian LU (2022) 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2022
Abstract
We introduce DiffPhy, a differentiable physics-based model for articulated 3d human motion reconstruction from video. Applications of physics-based reasoning in human motion analysis have so far been limited, both by the complexity of constructing adequate physical models of articulated human motion, and by the formidable challenges of performing stable and efficient inference with physics in the loop. We jointly address such modeling and inference challenges by proposing an approach that combines a physically plausible body representation with anatomical joint limits, a differentiable physics simulator, and optimization techniques that ensure good performance and robustness to suboptimal local optima. In contrast to several recent methods... (More)
We introduce DiffPhy, a differentiable physics-based model for articulated 3d human motion reconstruction from video. Applications of physics-based reasoning in human motion analysis have so far been limited, both by the complexity of constructing adequate physical models of articulated human motion, and by the formidable challenges of performing stable and efficient inference with physics in the loop. We jointly address such modeling and inference challenges by proposing an approach that combines a physically plausible body representation with anatomical joint limits, a differentiable physics simulator, and optimization techniques that ensure good performance and robustness to suboptimal local optima. In contrast to several recent methods [39], [42], [55], our approach readily supports full-body contact including interactions with objects in the scene. Most importantly, our model connects end-to-end with images, thus supporting direct gradient-based physics optimization by means of image-based loss functions. We validate the model by demonstrating that it can accurately reconstruct physically plausible 3d human motion from monocular video, both on public benchmarks with available 3d ground-truth, and on videos from the internet. (Less)
Please use this url to cite or link to this publication:
author
; ; and
organization
publishing date
type
Chapter in Book/Report/Conference proceeding
publication status
published
subject
host publication
Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition
publisher
IEEE - Institute of Electrical and Electronics Engineers Inc.
conference name
2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2022
conference location
New Orleans, United States
conference dates
2022-06-19 - 2022-06-24
external identifiers
  • scopus:85132486279
ISBN
978-1-6654-6946-3
978-1-6654-6947-0
DOI
10.1109/CVPR52688.2022.01284
language
English
LU publication?
yes
id
0b0162d1-762b-4d78-bc10-fa8373b7af80
date added to LUP
2022-05-06 10:49:05
date last changed
2024-07-11 19:45:56
@inproceedings{0b0162d1-762b-4d78-bc10-fa8373b7af80,
  abstract     = {{We introduce DiffPhy, a differentiable physics-based model for articulated 3d human motion reconstruction from video. Applications of physics-based reasoning in human motion analysis have so far been limited, both by the complexity of constructing adequate physical models of articulated human motion, and by the formidable challenges of performing stable and efficient inference with physics in the loop. We jointly address such modeling and inference challenges by proposing an approach that combines a physically plausible body representation with anatomical joint limits, a differentiable physics simulator, and optimization techniques that ensure good performance and robustness to suboptimal local optima. In contrast to several recent methods [39], [42], [55], our approach readily supports full-body contact including interactions with objects in the scene. Most importantly, our model connects end-to-end with images, thus supporting direct gradient-based physics optimization by means of image-based loss functions. We validate the model by demonstrating that it can accurately reconstruct physically plausible 3d human motion from monocular video, both on public benchmarks with available 3d ground-truth, and on videos from the internet.}},
  author       = {{Gärtner, Erik and Andriluka, Mykhaylo and Coumans, Erwin and Sminchisescu, Cristian}},
  booktitle    = {{Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition}},
  isbn         = {{978-1-6654-6946-3}},
  language     = {{eng}},
  publisher    = {{IEEE - Institute of Electrical and Electronics Engineers Inc.}},
  title        = {{Differentiable Dynamics for Articulated 3d Human Motion Reconstruction}},
  url          = {{http://dx.doi.org/10.1109/CVPR52688.2022.01284}},
  doi          = {{10.1109/CVPR52688.2022.01284}},
  year         = {{2022}},
}