Research map / Point clouds / Shape and surface completion
Intrinsic Point Cloud Interpolation via Dual Latent Space Navigation
Marie-Julie Rakotosaona, Maks Ovsjanikov
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 |
|---|---|---|---|
| interpolate in edge-length latent space and decode with shape decoder (dual latent space) instead of linear interpolation in PointNet AE latent space | PointNet AE · DFAUST interpolation | edge length variance: 0.3510 -> 0.2311 | Improved |
| dual latent space instead of PointNet AE on animals | PointNet AE · SMAL interpolation | edge length variance: 2.068 -> 1.538 | Improved |
| dual latent space in unsupervised setting | PointNet AE (unsupervised) · Unsupervised DFAUST interpolation | edge length variance: 0.597 -> 0.398 | Improved |
| add mapping networks to shape AE for reconstruction | PointNet AE · DFAUST reconstruction | edge length error (1e-5): 3.023 -> 1.641 | Improved |
| add mapping networks to shape AE for reconstruction | PointNet AE · DFAUST reconstruction | point cloud error (1e-4): 2.120 -> 2.572 | Hurt |
| add mapping networks for reconstruction on unseen dataset | Shape AE · SCAPE reconstruction | Chamfer distance (1e-3): 4.703 -> 4.135 | Improved |
| add cycle-consistency losses instead of direct reconstruction losses (ablation) | mapping networks · DFAUST reconstruction | edge length error: 0.1019 -> 1.641*10^{-5} | Improved |
| add L_map1 (rotation-invariant shape loss) to L_map2, L_map3 (ablation) | mapping networks · DFAUST reconstruction | point cloud error PC (1e-4): 14.816 -> 2.572 | Improved |
| add L_map2 to L_map1, L_map3 (ablation) | mapping networks · DFAUST reconstruction | edge length error EL (1e-5): 2.301 -> 1.641 | Improved |
| add L_map1 to L_map2, L_map3 effect on edge error (ablation) | mapping networks · DFAUST reconstruction | edge length error EL (1e-5): 1.595 -> 1.641 | No clear effect |
| add linearity regularization L_lin to edge auto-encoder (ablation) | edge auto-encoder · edge AE latent interpolation | edge length variance: 1.777 -> 0.199 | Improved |
| add linearity regularization L_lin to edge auto-encoder (ablation) | full method · DFAUST interpolation | edge length variance: 0.245 -> 0.230 | Improved |
Closest papers by meaning
- Neural Intrinsic Embedding for Non-Rigid Point Cloud Matching (CVPR 2023 · similarity 0.89)
- Meshing Point Clouds with Predicted Intrinsic-Extrinsic Ratio Guidance (ECCV 2020 · similarity 0.89)
- Learning a Structured Latent Space for Unsupervised Point Cloud Completion (CVPR 2022 · similarity 0.89)
- EditVAE: Unsupervised Parts-Aware Controllable 3D Point Cloud Shape Generation (AAAI 2022 · similarity 0.89)
- Self-Supervised Learning for Multimodal Non-Rigid 3D Shape Matching (CVPR 2023 · similarity 0.88)
- Digging into Intrinsic Contextual Information for High-fidelity 3D Point Cloud Completion (AAAI 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.
