Research map / Point clouds / Registration and matching
Neural Intrinsic Embedding for Non-Rigid Point Cloud Matching
Puhua Jiang, Mingze Sun, Ruqi Huang
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 bijectivity loss to the relative geodesic loss (ablation) | NIE · SCAPE_r embedding evaluation | OPT: 4.4 -> 3.5 | Improved |
| Add the KL loss to the relative geodesic and bijectivity losses (ablation) | NIE · SCAPE_r embedding evaluation | OPT: 3.5 -> 3.3 | Improved |
| Use modified DGCNN sampling aggregation (ablation) | NIE · SCAPE_r embedding evaluation | OPT: 3.3 -> 3.1 | Improved |
| Add the bijectivity loss (ablation) | NIE · SCAPE_r embedding evaluation | relative geodesic error: 8.8 -> 12.4 | Hurt |
| Add the KL loss after the relative geodesic and bijectivity losses (ablation) | NIE · SCAPE_r embedding evaluation | relative geodesic error: 12.4 -> 10.6 | Improved |
| Add bijectivity and KL losses to the relative geodesic loss (ablation) | NIE · final matching on SCAPE_r | matching error: 13.2 -> 11.5 | Improved |
| Increase basis dimension from 10 to 20 (ablation) | NIE · basis dimension ablation | OPT: 4.7 -> 3.1 | Improved |
| Increase basis dimension from 20 to 30 (ablation) | NIE · basis dimension ablation | relative geodesic error: 9.5 -> 11.3 | Hurt |
| Train with relative rather than absolute geodesic error | NIE embedding · SCAPE_r test shapes | relative geodesic error: 12.0 -> 9.5 | Improved |
| Train on SURREAL_r and test on re-meshed FAUST_r | NIE + NIM · FAUST_r matching | mean geodesic error: 15. -> 6.5 | Improved |
| Add Gaussian noise to test point clouds | NIE + NIM · FAUST_r matching trained on SURREAL_r | mean geodesic error: 6.5 -> 7.2 | Hurt |
| Match partial point clouds with holes | NIE + NIM · partial point cloud matching: hole | mean geodesic error: 15. -> 7.0 | Improved |
Closest papers by meaning
- Correspondence learning via linearly-invariant embedding (NeurIPS 2020 · similarity 0.91)
- Non-Rigid Shape Registration via Deep Functional Maps Prior (NeurIPS 2023 · similarity 0.90)
- Intrinsic Point Cloud Interpolation via Dual Latent Space Navigation (ECCV 2020 · similarity 0.89)
- PosDiffNet: Positional Neural Diffusion for Point Cloud Registration in a Large Field of View with Perturbations (AAAI 2024 · similarity 0.89)
- Occlusion-aware Non-Rigid Point Cloud Registration via Unsupervised Neural Deformation Correntropy (ICLR 2025 · similarity 0.89)
- DV-Matcher: Deformation-based Non-rigid Point Cloud Matching Guided by Pre-trained Visual Features (CVPR 2025 · similarity 0.89)
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.
