Research map / 3D vision / Face and shape
3D-CODED: 3D Correspondences by Deep Deformation
Thibault Groueix, Matthew Fisher, Vladimir G. Kim, Bryan C. Russell, Mathieu Aubry
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 latent-feature reconstruction optimization (regression) step (ablation) | 3D-CODED · FAUST-inter | error (cm): 6.29 -> 3.255 | Improved |
| add regular point sampling on training surfaces (ablation) | 3D-CODED with regression · FAUST-inter | error (cm): 3.255 -> 3.048 | Improved |
| add high-resolution template for the nearest-neighbor step (ablation) | 3D-CODED · FAUST-inter | error (cm): 3.048 -> 2.878 | Improved |
| train on augmented synthetic data instead of FAUST training set (ablation) | 3D-CODED · FAUST-inter | error (cm): 18.22 -> 3.26 | Improved |
| augment synthetic dataset over the non-augmented one (ablation) | 3D-CODED · FAUST-inter | error (cm): 5.63 -> 3.26 | Improved |
| increase augmented training shapes from 1e3 to 2.3e5 (ablation) | 3D-CODED · FAUST-inter | error (cm): 5.76 -> 3.26 | Improved |
| add correspondence supervision over regularized unsupervised training (ablation) | 3D-CODED · FAUST-inter | error (cm): 4.835 -> 2.878 | Improved |
| add surface regularization to the Chamfer-only unsupervised loss (ablation) | 3D-CODED · FAUST-inter | error (cm): 8.727 -> 4.835 | Improved |
| use FAUST template instead of zero or separated template (ablation) | 3D-CODED · FAUST-inter | error (cm): 3.385 -> 3.255 (zero), 3.314 -> 3.255 (separated) | Improved |
| use symmetric Chamfer distance instead of asymmetric (ablation) | 3D-CODED · FAUST-inter | error (cm): 4.023 -> 3.255, 3.336 -> 3.255 | Improved |
| use rotation-invariant network instead of orientation search (ablation) | 3D-CODED · FAUST-inter | error (cm): 2.878 -> 3.10 | Hurt |
| increase TOSCA sampling perturbation to strongest level | 3D-CODED · TOSCA horse | error (cm): 4.82 -> 9.93 | Hurt |
Later papers that compared with this method
15 later accepted papers report 3DCODED as a baseline in their results tables; 12 of them report a better number on most of the results they share with it.
- PointHPS: Cascaded 3D Human Pose and Shape Estimation from Point Clouds (IJCV 2026 · better on 2 of 2 shared results)
- Basis Restricted Elastic Shape Analysis on the Space of Unregistered Surfaces (IJCV 2024 · better on 7 of 8 shared results)
- Self-supervised Learning of Implicit Shape Representation with Dense Correspondence for Deformable Objects (ICCV 2023 · better on 7 of 7 shared results)
- BaRe-ESA: A Riemannian Framework for Unregistered Human Body Shapes (ICCV 2023 · better on 3 of 4 shared results)
- Implicit Field Supervision for Robust Non-rigid Shape Matching (ECCV 2022 · better on 1 of 1 shared result)
- Unsupervised Dense Deformation Embedding Network for Template-Free Shape Correspondence (ICCV 2021 · better on 5 of 5 shared results)
- Recurrent Multi-View Alignment Network for Unsupervised Surface Registration (CVPR 2021 · better on 1 of 1 shared result)
- Neural Feature Matching in Implicit 3D Representations (ICML 2021 · better on 0 of 1 shared result)
7 more not listed.
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
- Unsupervised Learning of Dense Shape Correspondence (CVPR 2019 · similarity 0.91)
- Learning elementary structures for 3D shape generation and matching (NeurIPS 2019 · similarity 0.91)
- Deformed Implicit Field: Modeling 3D Shapes With Learned Dense Correspondence (CVPR 2021 · similarity 0.90)
- Reduced Representation of Deformation Fields for Effective Non-rigid Shape Matching (NeurIPS 2022 · similarity 0.90)
- Unsupervised Dense Deformation Embedding Network for Template-Free Shape Correspondence (ICCV 2021 · similarity 0.90)
- CorrNet3D: Unsupervised End-to-End Learning of Dense Correspondence for 3D Point Clouds (CVPR 2021 · similarity 0.90)
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.
