Research map / 3D vision / Face and shape
PAUL: Procrustean Autoencoder for Unsupervised Lifting
Chaoyang Wang, Simon Lucey
Check your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use the Procrustean auto-encoder prior instead of the compared NRSfM methods | PAUL · CMU motion capture S1, S5, S64, S70, S123 | normalized error: 10.74 -> 4.97; 13.40 -> 4.38; 4.38 -> 0.39; 2.17 -> 0.77; 2.23 -> 0.59 | Improved |
| Use the Procrustean auto-encoder prior instead of the strongest listed baseline | PAUL · Pascal3D+ mean and 8 classes | normalized error (%): 10.3 -> 8.4; 7.58 -> 5.32 | Improved |
| Use PAUL instead of Deep NRSfM++ | PAUL · UP3D 79KP test | MPJPE: 0.062 -> 0.058 | Improved |
| Use PAUL instead of PRN | PAUL · SURREAL test | normalized error (%): 13.77 -> 12.36 | Improved |
| Use PAUL instead of PRN | PAUL · H3.6M validation | MPJPE: 86.4 -> 88.3; 124.5 -> 132.5 | Hurt |
| Add the auto-encoder reconstruction term to the decoder-only approach (ablation) | PAUL · CMU motion capture | reconstruction error | Improved |
| Add a low-rank constraint to the decoder-only approach (ablation) | ADL + low rank · CMU motion capture | reconstruction error | Mixed |
| Vary bottleneck dimension from 2 to 12 (ablation) | PAUL · NRSfM datasets | normalized error | No clear effect |
| Use a larger bottleneck dimension on noisy data (ablation) | PAUL · CMU motion capture with Gaussian keypoint noise | reconstruction accuracy | Hurt |
| Use the auto-encoder objective rather than decoder-only learning (ablation) | PAUL versus ADL · Latent-space visualization on motion sequences | latent trajectory continuity | Improved |
Later papers that compared with this method
4 later accepted papers report PAUL as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.
- Deep Non-Rigid Structure-from-Motion Revisited: Canonicalization and Sequence Modeling (AAAI 2025 · better on 1 of 1 shared result)
- Deep Non-Rigid Structure-From-Motion: A Sequence-to-Sequence Translation Perspective (TPAMI 2024 · better on 1 of 1 shared result)
- Neural Prior for Trajectory Estimation (CVPR 2022 · better on 5 of 5 shared results)
- MHR-Net: Multiple-Hypothesis Reconstruction of Non-rigid Shapes from 2D Views (ECCV 2022 · better on 2 of 2 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- C3DPO — C3DPO: Canonical 3D Pose Networks for Non-Rigid Structure From Motion (ICCV 2019) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Procrustean Regression Networks: Learning 3D Structure of Non-Rigid Objects from 2D Annotations (ECCV 2020 · similarity 0.90)
- 3D-LFM: Lifting Foundation Model (CVPR 2024 · similarity 0.87)
- Neural Dense Non-Rigid Structure from Motion with Latent Space Constraints (ECCV 2020 · similarity 0.87)
- Deep Non-Rigid Structure-From-Motion: A Sequence-to-Sequence Translation Perspective (TPAMI 2024 · similarity 0.87)
- Deep Non-Rigid Structure-from-Motion Revisited: Canonicalization and Sequence Modeling (AAAI 2025 · similarity 0.86)
- Non-Rigid Structure-from-Motion: Temporally-Smooth Procrustean Alignment and Spatially-Variant Deformation Modeling (CVPR 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
