AI Paper Atlas

Research map / Point clouds / Point cloud classification

A Statistical Manifold Framework for Point Cloud Data

Yonghyeon Lee, Seungyeon Kim, Jinwon Choi, Frank Park

ICML 2022 · Point clouds › Point cloud classification · approach: Local feature aggregation networks

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
add info-Riemannian metric regularization (vs. no regularization)FcNet · ModelNet40 transfer from ShapeNetclassification accuracy: +2.1 (88.3% to 90.4%)Improved
add info-Riemannian metric regularization (vs. no regularization)FoldingNet · ModelNet40 transfer from ShapeNetclassification accuracy: +0.8 (89.3% to 90.1%)Improved
add info-Riemannian metric regularization (vs. no regularization)PointCapsNet · ModelNet40 transfer from ShapeNetclassification accuracy: +1.3 (87.2% to 88.5%)Improved
add info-Riemannian metric regularization (vs. no regularization)DGCNN-FcNet · ModelNet40 transfer from ShapeNetclassification accuracy: +0.7 (90.3% to 91.0%)Improved
add Euclidean-metric regularization vs. info-Riemannian regularization (ablation)FcNet, FoldingNet, PointCapsNet, DGCNN-FcNet · ModelNet40 transfer from ShapeNetclassification accuracy: FcNet+E 89.3 vs +I 90.4; FoldingNet+E 88.9 vs +I 90.1; PointCapsNet+E 88.1 vs +I 88.5; DGCNN-FcNet+E 89.9 vs +I 91.0Mixed
add info-Riemannian metric regularization (vs. no regularization)FcNet · ModelNet10 transfer from ShapeNetclassification accuracy: +0.8 (93.5% to 94.3%)Improved
add info-Riemannian metric regularization (vs. no regularization)FoldingNet · ModelNet10 transfer from ShapeNetclassification accuracy: +0.8 (93.7% to 94.5%)Improved
add info-Riemannian metric regularization (vs. no regularization)DGCNN-FcNet · ModelNet10 transfer from ShapeNetclassification accuracy: +0.7 (94.5% to 95.2%)Improved
add info-Riemannian metric regularization under 20% input noiseFcNet · ModelNet40 transfer, 20% noiseclassification accuracy: +7.9 (64.5% to 72.4%)Improved
add Euclidean-metric regularization under 20% input noise (ablation)FcNet · ModelNet40 transfer, 20% noiseclassification accuracy: +5.9 (64.5% to 70.4%)Improved
add info-Riemannian metric regularization under 1% label rateFcNet · ModelNet40 semi-supervised, 1% labelsclassification accuracy: +7.1 (50.3% to 57.4%)Improved
add info-Riemannian metric regularization on synthetic shapeslatent space of point cloud autoencoder · Synthetic shapes (27 datasets), GMM clusteringNMI and ARI: NMI 0.7624 to 0.9484; ARI 0.7209 to 0.9368Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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