Research map / 3D vision / Face and shape
Procrustean Regression Networks: Learning 3D Structure of Non-Rigid Objects from 2D Annotations
Sungheon Park, Minsik Lee, Nojun Kwak
Read the paperCheck 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 |
|---|---|---|---|
| add the Procrustean low-rank regularization term (ablation) | PRN-FCN · Human3.6M 2D-input (GT-ortho) | MPJPE: 144.8 -> 66.7 | Improved |
| add the Procrustean low-rank regularization term (ablation) | PRN-CNN · Human3.6M RGB input | MPJPE: 164.5 -> 108.9 | Improved |
| add the Procrustean low-rank regularization term (ablation) | PRN-FCN · 300-VW 2D-input | normalized error: 0.5201 -> 0.1512 | Improved |
| add the Procrustean low-rank regularization term (ablation) | PRN-FCN · SURREAL 2D-input | normalized error: 0.3565 -> 0.1377 | Improved |
| weight reprojection error by keypoint detection scores | PRN-FCN · Human3.6M 2D-input (SH fine-tuned) | MPJPE: 99.1 -> 95.9 | Improved |
| weight reprojection error by keypoint scores | PRN-FCN · Human3.6M Sitting (SH FT) | MPJPE: 114.0 -> 102.1 | Improved |
| weight reprojection error by keypoint scores | PRN-FCN · Human3.6M SittingDown (SH FT) | MPJPE: 141.2 -> 113.2 | Improved |
| fine-tune the 2D keypoint detector on Human3.6M | PRN-FCN · Human3.6M 2D-input (SH) | MPJPE: 124.5 -> 99.1 | Improved |
| use PRN instead of C3DPO under perspective projection | PRN-FCN · Human3.6M 2D-input (GT-persp) | MPJPE: 95.6 -> 86.4 | Improved |
| use PRN instead of C3DPO under orthographic projection | PRN-FCN · Human3.6M 2D-input (GT-ortho) | MPJPE: 67.8 -> 66.7 | No clear effect |
| train on PRN loss instead of NRSfM-reconstructed labels (SPM) | PRN-FCN · Human3.6M 2D-input (GT-ortho) | MPJPE: 82.3 -> 66.7 | Improved |
| increase ratio of missing input points from 0% to 20% (ablation) | PRN-FCN · Human3.6M 2D-input | MPJPE: 66.7 -> 73.8 | Hurt |
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 5 of 5 shared results
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
- Neural Dense Non-Rigid Structure from Motion with Latent Space Constraints (ECCV 2020 · similarity 0.91)
- Deep Non-Rigid Structure-from-Motion Revisited: Canonicalization and Sequence Modeling (AAAI 2025 · similarity 0.90)
- Non-Rigid Structure-from-Motion: Temporally-Smooth Procrustean Alignment and Spatially-Variant Deformation Modeling (CVPR 2024 · similarity 0.89)
- REACTO: Reconstructing Articulated Objects from a Single Video (CVPR 2024 · similarity 0.87)
- TFS-NeRF: Template-Free NeRF for Semantic 3D Reconstruction of Dynamic Scene (NeurIPS 2024 · similarity 0.86)
- Learning Implicit Representation for Reconstructing Articulated Objects (ICLR 2024 · similarity 0.86)
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.
