Research map / 3D vision / Face and shape
Self-supervised Learning of Implicit Shape Representation with Dense Correspondence for Deformable Objects
Baowen Zhang, Jiahe Li, Xiaoming Deng, Yinda Zhang, Cuixia Ma, Hongan Wang
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 neighborhood rigid loss L_nbr (ablation) | Ours · DeformingThings4D bear | Chamfer distance: 2.045 -> 0.439 | Improved |
| add the self-correspondence loss L_recon (ablation) | Ours · DeformingThings4D bear | Chamfer distance: 358.129 -> 0.439 | Improved |
| add the piece-wise rigid loss L_pr (ablation) | Ours · DeformingThings4D bear | correspondence error: 0.0858 -> 0.0700 | Improved |
| add the reflection penalty on top of L_arap in local rigid loss (ablation) | Ours · DeformingThings4D bear | Chamfer distance: 0.609 -> 0.439 | Improved |
| use proposed local rigid loss instead of elastic loss (ablation) | Ours · DeformingThings4D bear | Chamfer distance: 0.815 -> 0.439 | Improved |
| add template normal loss L_pfn (ablation) | Ours · DeformingThings4D bear | Chamfer distance: 0.547 -> 0.439 | Improved |
| add SDF value as encoder input (ablation) | Ours · D-FAUST subject 50026 | correspondence error: 0.0265 -> 0.0141 | Improved |
| add part prediction network psi_e (ablation) | Ours · D-FAUST subject 50026 | Chamfer distance: 1.174 -> 0.687 | Improved |
| increase number of parts from 5 to 20 (ablation) | piece-wise rigid constraint · D-FAUST | correspondence error: 0.0448 -> 0.0141 | Improved |
| neighborhood sigma 0.05 versus 0.1 (ablation) | neighborhood rigid constraint · D-FAUST | correspondence error: 0.0520 -> 0.0141 | Improved |
| embed template in the shape latent space instead of a separate template (ablation) | Ours · qualitative template visualization | template quality | Improved |
| use the proposed method versus 3D-CODED | DeformingThings4D · DeformingThings4D average | Chamfer distance: 1.658 -> 1.165 | Mixed |
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.
- 3D-CODED — 3D-CODED: 3D Correspondences by Deep Deformation (ECCV 2018) · better on 7 of 7 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
- Semantic-Aware Implicit Template Learning via Part Deformation Consistency (ICCV 2023 · similarity 0.91)
- Deformed Implicit Field: Modeling 3D Shapes With Learned Dense Correspondence (CVPR 2021 · similarity 0.91)
- Neural Pose Representation Learning for Generating and Transferring Non-Rigid Object Poses (NeurIPS 2024 · similarity 0.90)
- Unsupervised Learning of Dense Shape Correspondence (CVPR 2019 · similarity 0.90)
- Learning SO(3)-Invariant Semantic Correspondence via Local Shape Transform (CVPR 2024 · similarity 0.88)
- Implicit Modeling of Non-rigid Objects with Cross-Category Signals (AAAI 2024 · similarity 0.88)
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.
