Research map / Point clouds / Shape and surface completion
DynoSurf: Neural Deformation-based Temporally Consistent Dynamic Surface Reconstruction
Yuxin Yao, Siyu Ren, Junhui Hou, Zhi Deng, Juyong Zhang, Wenping 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 |
|---|---|---|---|
| replace learnable control-point deformation with pointwise MLP deformation (ablation) | DynoSurf · AMA ablation | Corr. error (x10^-2): 2.89 -> 5.91 | Hurt |
| learn blending weights instead of fixed Gaussian formula (ablation) | DynoSurf · AMA ablation | CD (x10^-4): 0.475 -> 0.324 | Improved |
| make control point positions learnable (ablation) | DynoSurf · AMA ablation | CD (x10^-4): 0.372 -> 0.324 | Improved |
| add normal consistency loss in temporal reconstruction (ablation) | DynoSurf · AMA ablation | CD (x10^-4): 0.656 -> 0.324 | Improved |
| add deformation smoothness loss (ablation) | DynoSurf · AMA ablation | CD (x10^-4): 0.378 -> 0.324 | Improved |
| add robust SDF loss (ablation) | DynoSurf · AMA ablation | CD (x10^-4): CD 0.330 -> 0.324, Corr. 3.14 -> 2.89 | Mixed |
| add shape preservation loss (ablation) | DynoSurf · AMA ablation | CD (x10^-4): 0.335 -> 0.324 (Corr. 2.91 -> 2.89) | Mixed |
| increase control points from 10 to 30 (ablation) | DynoSurf · AMA ablation | CD (x10^-4): 0.437 -> 0.324 | Improved |
| increase control points from 30 to 100 (ablation) | DynoSurf · AMA ablation | Corr. error (x10^-2): CD 0.324 -> 0.315, Corr. 2.89 -> 3.16 | Mixed |
| use ground-truth normals instead of estimated normals (ablation) | DynoSurf · AMA | CD (x10^-4): 0.324 -> 0.318 | No clear effect |
| increase input sequence length from 30 to 80 frames (ablation) | DynoSurf · DFAUST shake_shoulders | CD (x10^-4): 0.138 -> 0.265 | Hurt |
| reduce input points per frame from 5000 to 2000 (ablation) | DynoSurf · AMA | CD (x10^-4): 0.324 -> 0.451 | Hurt |
Later papers that compared with this method
1 later accepted paper reports DynoSurf as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Neu-PiG: Neural Preconditioned Grids for Fast Dynamic Surface Reconstruction on Long Sequences (CVPR 2026 · 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
- Neu-PiG: Neural Preconditioned Grids for Fast Dynamic Surface Reconstruction on Long Sequences (CVPR 2026 · similarity 0.89)
- DySurface: Consistent 4D Surface Reconstruction via Bridging Explicit Gaussians and Implicit Functions (NeurIPS 2026 · similarity 0.89)
- Dynamic Neural Surfaces for Elastic 4D Shape Representation and Analysis (CVPR 2025 · similarity 0.88)
- 4Deform: Neural Surface Deformation for Robust Shape Interpolation (CVPR 2025 · similarity 0.88)
- Diffeomorphic Mesh Deformation via Efficient Optimal Transport for Cortical Surface Reconstruction (ICLR 2024 · similarity 0.88)
- Implicit Neural Surface Deformation with Explicit Velocity Fields (ICLR 2025 · similarity 0.87)
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.
