AI Paper Atlas

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

ICCV 2021 · Point clouds › Point cloud classification · approach: Local feature aggregation networks · 76 citations (OpenAlex)

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 changedWhereMeasuredEffect
train with sliced Wasserstein loss instead of ChamferPointNet autoencoder · ModelNet40 classificationaccuracy (%): 83.9 -> 86.8Improved
train with sliced Wasserstein loss instead of EMDPointNet autoencoder · ModelNet40 classificationaccuracy (%): 84.4 -> 86.8Improved
train with sliced Wasserstein loss instead of ChamferPointNet autoencoder · ModelNet40 reconstructionChamfer discrepancy: 0.014 -> 0.007Improved
train with sliced Wasserstein loss instead of ChamferPointNet autoencoder · ModelNet40 reconstructionEMD: 0.314 -> 0.091Improved
train with sliced Wasserstein loss instead of ChamferPointNet autoencoder · 3DMatch registrationrecall (%): 51.3 -> 57.4Improved
use sliced Wasserstein loss instead of ChamferPoint cloud capsule network · ModelNet40 classificationaccuracy (%): 88.45 -> 88.78Improved
use sliced Wasserstein loss instead of EMDautoencoder training · Training time per iterationruntime (ms): 385 -> 138Improved
use sliced Wasserstein loss instead of Chamferautoencoder training · Convergence of exact EMD error at epoch 300EMD error: 0.314 -> 0.091Improved
increase number of Monte Carlo slices from 1 to 100 (ablation)SSW-AE · ModelNet40 classificationaccuracy (%): 86.5 -> 86.8No clear effect
increase number of Monte Carlo slices from 1 to 50 (ablation)SSW-AE · 3DMatch registrationrecall (%): 55.1 -> 57.7Improved
use sliced Wasserstein loss instead of EMD for generator latent spaceSSW-AE · ShapeNet chair generationJSD x10^2 / MMD-CD: 3.73 -> 3.24; 0.61 -> 0.79Mixed
use max-sliced instead of sliced Wasserstein lossMSW-AE · ShapeNet chair generationMMD-EMD: 11.22 -> 9.64Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy