mlsanityData to 7 Oct 2026

PAUL: Procrustean Autoencoder for Unsupervised Lifting

arXiv 2103.16773 ↗First posted 31 Mar 2021cs.CVcited by 9

Not on the map. Why?

Abstract (arXiv)

Recent success in casting Non-rigid Structure from Motion (NRSfM) as an unsupervised deep learning problem has raised fundamental questions about what novelty in NRSfM prior could the deep learning offer. In this paper we advocate for a 3D deep auto-encoder framework to be used explicitly as the NRSfM prior. The framework is unique as: (i) it learns the 3D auto-encoder weights solely from 2D projected measurements, and (ii) it is Procrustean in that it jointly resolves the unknown rigid pose for each shape instance. We refer to this architecture as a Procustean Autoencoder for Unsupervised Lifting (PAUL), and demonstrate state-of-the-art performance across a number of benchmarks in comparison to recent innovations such as Deep NRSfM and C3PDO.

Key facts. Cited by 9 papers in the map (data through 7 Oct 2026). Builds on 17 of its 44 references that are in the map. First posted on arXiv 31 Mar 2021.

Builds on

17 of 44 references.

References · 17

  1. 2014Adam: A Method for Stochastic Optimizationcited by 36,960
  2. 2017A simple yet effective baseline for 3d human pose estimationcited by 401
  3. 2017Learning from Synthetic Humanscited by 358
  4. 2017Unite the People: Closing the Loop Between 3D and 2D Human Representationscited by 205
  5. 2019From Variational to Deterministic Autoencoderscited by 131
  6. 2019Unsupervised 3D Pose Estimation with Geometric Self-Supervisioncited by 95
  7. 2019RepNet: Weakly Supervised Training of an Adversarial Reprojection Network for 3D Human Pose Estimationcited by 85
  8. 2019C3DPO: Canonical 3D Pose Networks for Non-Rigid Structure From Motioncited by 60
  9. 2015Sparseness Meets Deepness: 3D Human Pose Estimation from Monocular Videocited by 54
  10. 2018Can 3D Pose be Learned from 2D Projections Alone?cited by 44
  11. 2019Deep Non-Rigid Structure from Motioncited by 38
  12. 2018Unsupervised Adversarial Learning of 3D Human Pose from 2D Joint Locationscited by 18
5 more
  1. 2019Non-Rigid Structure from Motion: Prior-Free Factorization Method Revisitedcited by 18
  2. 2016Multi-body Non-rigid Structure-from-Motioncited by 13
  3. 2020Procrustean Regression Networks: Learning 3D Structure of Non-Rigid Objects from 2D Annotationscited by 13
  4. 2019Variational Auto-Decoder: A Method for Neural Generative Modeling from Incomplete Datacited by 10
  5. 2020Deep NRSfM++: Towards Unsupervised 2D-3D Lifting in the Wildcited by 10

Built on by

Cited by year

Citing papers per year: 2021: 1; 2022: 5; 2023: 1; 2024: 1; 2025: 1.

9 papers.

  1. 2022State of the Art in Dense Monocular Non-Rigid 3D Reconstructioncited by 30
  2. 2022Unbiased 4D: Monocular 4D Reconstruction with a Neural Deformation Modelcited by 25
  3. 2022Deep Non-rigid Structure-from-Motion: A Sequence-to-Sequence Translation Perspectivecited by 8
  4. 2022MHR-Net: Multiple-Hypothesis Reconstruction of Non-Rigid Shapes from 2D Viewscited by 7
  5. 2023Unsupervised 3D Pose Estimation with Non-Rigid Structure-from-Motion Modelingcited by 7
  6. 2021End-to-End Learning of Multi-category 3D Pose and Shape Estimationcited by 1
  7. 2025Non-Rigid Structure-from-Motion via Differential Geometry with Recoverable Conformal Scalecited by 1
  8. 2022Piecewise Planar Hulls for Semi-Supervised Learning of 3D Shape and Pose from 2D Imagescited by 0
  9. 2024Object Agnostic 3D Lifting in Space and Timecited by 0