Research map / Point clouds / Point cloud classification
Revisiting Point Cloud Simplification: A Learnable Feature Preserving Approach
Rolandos Alexandros Potamias, Giorgos Bouritsas, Stefanos Zafeiriou
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 adaptive (curvature-weighted) Chamfer distance instead of plain CD (ablation) | Proposed simplifier · TOSCA (ratio 0.1) | SDM (x10^-4): 9.64 -> 9.61 | Improved |
| add curvature preservation loss on top of adaptive CD (ablation) | Proposed simplifier · TOSCA (ratio 0.1) | SDM (x10^-4): 9.61 -> 7.84 | Improved |
| add curvature preservation loss on top of adaptive CD (ablation) | Proposed simplifier · TOSCA (ratio 0.1) | RE (x10^-4): 4.42 -> 4.41 | Improved |
| add adaptive CD instead of plain CD (ablation) | Proposed simplifier · TOSCA (ratio 0.1) | Chamfer distance: 2.41 -> 2.54 | Hurt |
| train on TOSCA instead of MeIn3D, tested zero-shot on ModelNet-like diverse shapes (ablation) | Proposed simplifier · MeIn3D vs TOSCA (ratio 0.1, TOSCA test) | Chamfer distance: 2.53 -> 2.45 | Improved |
| use proposed learned simplifier instead of QEM | Point cloud simplification · TOSCA (ratio 0.1) | MSDM (x10^-3): 10.4 -> 7.84 | Improved |
| use proposed learned simplifier instead of FPS | Point cloud simplification · TOSCA (ratio 0.1) | Chamfer distance: 1.93 -> 2.45 | Hurt |
| use proposed simplifier instead of spectral mesh method | Point cloud simplification · TOSCA (ratio 0.1) | Chamfer distance: 3.12 -> 2.45 | Improved |
| apply TOSCA-trained model zero-shot to ModelNet instead of training on ModelNet | Proposed simplifier · ModelNet (ratio 0.5) | MSDM (x10^-3): 1.28 -> 1.42 | Hurt |
| apply proposed method to Gaussian-noised inputs instead of clean | Proposed simplifier · TOSCA (ratio 0.1) | SDM (x10^-4): 7.84 -> 14.42 | Hurt |
| use proposed method instead of FPS on noisy inputs | Point cloud simplification · TOSCA with Gaussian noise (ratio 0.1) | SDM (x10^-4): 15.86 -> 14.42 | Improved |
| use proposed method instead of FPS on real LiDAR scans | Point cloud simplification · Toronto3D (ratio 0.1) | SDM (x10^-4): 0.46 -> 0.38 | Improved |
Closest papers by meaning
- Learning Point Cloud Completion without Complete Point Clouds: A Pose-Aware Approach (ICCV 2023 · similarity 0.86)
- Rethinking the compositionality of point clouds through regularization in the hyperbolic space (NeurIPS 2022 · similarity 0.84)
- RegGeoNet: Learning Regular Representations for Large-Scale 3D Point Clouds (IJCV 2022 · similarity 0.84)
- LTA-PCS: Learnable Task-Agnostic Point Cloud Sampling (CVPR 2024 · similarity 0.83)
- ComPC: Completing a 3D Point Cloud with 2D Diffusion Priors (ICLR 2025 · similarity 0.83)
- SAMBLE: Shape-Specific Point Cloud Sampling for an Optimal Trade-Off Between Local Detail and Global Uniformity (CVPR 2025 · similarity 0.83)
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.
