Research map / Point clouds / Point cloud classification
Point-Set Distances for Learning Representations of 3D Point Clouds
Trung Nguyen, Quang-Hieu Pham, Tam Le, Tung Pham, Nhat Ho, Binh-Son Hua
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 |
|---|---|---|---|
| train with sliced Wasserstein loss instead of Chamfer | PointNet autoencoder · ModelNet40 classification | accuracy (%): 83.9 -> 86.8 | Improved |
| train with sliced Wasserstein loss instead of EMD | PointNet autoencoder · ModelNet40 classification | accuracy (%): 84.4 -> 86.8 | Improved |
| train with sliced Wasserstein loss instead of Chamfer | PointNet autoencoder · ModelNet40 reconstruction | Chamfer discrepancy: 0.014 -> 0.007 | Improved |
| train with sliced Wasserstein loss instead of Chamfer | PointNet autoencoder · ModelNet40 reconstruction | EMD: 0.314 -> 0.091 | Improved |
| train with sliced Wasserstein loss instead of Chamfer | PointNet autoencoder · 3DMatch registration | recall (%): 51.3 -> 57.4 | Improved |
| use sliced Wasserstein loss instead of Chamfer | Point cloud capsule network · ModelNet40 classification | accuracy (%): 88.45 -> 88.78 | Improved |
| use sliced Wasserstein loss instead of EMD | autoencoder training · Training time per iteration | runtime (ms): 385 -> 138 | Improved |
| use sliced Wasserstein loss instead of Chamfer | autoencoder training · Convergence of exact EMD error at epoch 300 | EMD error: 0.314 -> 0.091 | Improved |
| increase number of Monte Carlo slices from 1 to 100 (ablation) | SSW-AE · ModelNet40 classification | accuracy (%): 86.5 -> 86.8 | No clear effect |
| increase number of Monte Carlo slices from 1 to 50 (ablation) | SSW-AE · 3DMatch registration | recall (%): 55.1 -> 57.7 | Improved |
| use sliced Wasserstein loss instead of EMD for generator latent space | SSW-AE · ShapeNet chair generation | JSD x10^2 / MMD-CD: 3.73 -> 3.24; 0.61 -> 0.79 | Mixed |
| use max-sliced instead of sliced Wasserstein loss | MSW-AE · ShapeNet chair generation | MMD-EMD: 11.22 -> 9.64 | Improved |
Closest papers by meaning
- Fast Estimation of Wasserstein Distances via Regression on Sliced Wasserstein Distances (ICLR 2026 · similarity 0.89)
- Towards Better Spherical Sliced-Wasserstein Distance Learning with Data-Adaptive Discriminative Projection Direction (AAAI 2025 · similarity 0.87)
- An in depth look at the Procrustes-Wasserstein distance: properties and barycenters (ICML 2025 · similarity 0.87)
- Stereographic Spherical Sliced Wasserstein Distances (ICML 2024 · similarity 0.87)
- LinNet: Linear Network for Efficient Point Cloud Representation Learning (NeurIPS 2024 · similarity 0.87)
- Sonata: Self-Supervised Learning of Reliable Point Representations (CVPR 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.
