Research map / 3D vision / Face and shape
3D Pose Transfer with Correspondence Learning and Mesh Refinement
Chaoyue Song, Jiacheng Wei, Ruibo Li, Fayao Liu, Guosheng Lin
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 |
|---|---|---|---|
| Use 3D-CoreNet instead of NPT | 3D-CoreNet · SMPL | PMD (x1e-3): 0.66 -> 0.08 | Improved |
| Use 3D-CoreNet instead of DT (needs key points) | 3D-CoreNet · SMPL | PMD (x1e-3): 0.15 -> 0.08 | Improved |
| Use 3D-CoreNet instead of NPT | 3D-CoreNet · SMAL | EMD (x1e-2): 11.65 -> 7.28 | Improved |
| Use 3D-CoreNet instead of DT | 3D-CoreNet · SMAL | PMD (x1e-3): 13.37 -> 2.26 | Improved |
| Remove mesh refinement, use correlation matrix only (ablation) | 3D-CoreNet · SMPL | PMD (x1e-3): 0.08 -> 0.46 | Hurt |
| Remove mesh refinement, use optimal matching matrix only (ablation) | 3D-CoreNet · SMPL | PMD (x1e-3): 0.08 -> 0.44 | Hurt |
| Use optimal matching matrix instead of correlation matrix (no refinement) (ablation) | 3D-CoreNet · SMPL | EMD (x1e-2): 3.49 -> 3.42 | Improved |
| Replace ElaIN with SPAdaIN (ablation) | 3D-CoreNet · SMPL | CD (x1e-3): 0.22 -> 0.37 | Hurt |
| Remove edge loss (ablation) | 3D-CoreNet · SMPL | EMD (x1e-2): 1.89 -> 2.28 | Hurt |
| Remove edge loss (ablation) | 3D-CoreNet · SMPL | PMD (x1e-3): 0.08 -> 0.14 | Hurt |
| Solve optimal transport (matching matrix) instead of correlation matrix at inference | 3D-CoreNet · SMPL | inference time: 0.0124s -> 0.0131s | No clear effect |
| Use 3D-CoreNet instead of DT | 3D-CoreNet · SMPL | inference time: 3.3352s -> 0.0131s | Improved |
Closest papers by meaning
- Unsupervised 3D Pose Transfer With Cross Consistency and Dual Reconstruction (TPAMI 2023 · similarity 0.91)
- Neural Pose Transfer by Spatially Adaptive Instance Normalization (CVPR 2020 · similarity 0.91)
- Weakly-supervised 3D Pose Transfer with Keypoints (ICCV 2023 · similarity 0.88)
- Towards Robust 3D Pose Transfer with Adversarial Learning (CVPR 2024 · similarity 0.87)
- Neural Pose Representation Learning for Generating and Transferring Non-Rigid Object Poses (NeurIPS 2024 · similarity 0.86)
- REACTO: Reconstructing Articulated Objects from a Single Video (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.
