Adversarial Parametric Pose Prior

Andrey Davydov, Anastasia Remizova, Victor Constantin, Sina Honari, Mathieu Salzmann, Pascal Fua

CVPR 2022 pp. 10997-11005

doi:10.1109/CVPR52688.2022.01072 /cvpr/2022/davydov2022cvpr-adversarial/

Abstract

The Skinned Multi-Person Linear (SMPL) model represents human bodies by mapping pose and shape parameters to body meshes. However, not all pose and shape parameter values yield physically-plausible or even realistic body meshes. In other words, SMPL is under-constrained and may yield invalid results. We propose learning a prior that restricts the SMPL parameters to values that produce realistic poses via adversarial training. We show that our learned prior covers the diversity of the real-data distribution, facilitates optimization for 3D reconstruction from 2D keypoints, and yields better pose estimates when used for regression from images. For all these tasks, it outperforms the state-of-the-art VAE-based approach to constraining the SMPL parameters. The code will be made available at https://github.com/cvlab-epfl/adv_param_pose_prior.

PDF CVPR Semantic Scholar

Cite

Text

Davydov et al. "Adversarial Parametric Pose Prior." Conference on Computer Vision and Pattern Recognition, 2022. doi:10.1109/CVPR52688.2022.01072

Markdown

[Davydov et al. "Adversarial Parametric Pose Prior." Conference on Computer Vision and Pattern Recognition, 2022.](https://mlanthology.org/cvpr/2022/davydov2022cvpr-adversarial/) doi:10.1109/CVPR52688.2022.01072

BibTeX

@inproceedings{davydov2022cvpr-adversarial,
  title     = {{Adversarial Parametric Pose Prior}},
  author    = {Davydov, Andrey and Remizova, Anastasia and Constantin, Victor and Honari, Sina and Salzmann, Mathieu and Fua, Pascal},
  booktitle = {Conference on Computer Vision and Pattern Recognition},
  year      = {2022},
  pages     = {10997-11005},
  doi       = {10.1109/CVPR52688.2022.01072},
  url       = {https://mlanthology.org/cvpr/2022/davydov2022cvpr-adversarial/}
}