AI Paper Atlas

Research map / Point clouds / Point cloud classification

A Closer Look at Local Aggregation Operators in Point Cloud Analysis

Ze Liu(†), Han Hu, Yue Cao, Zheng Zhang, Xin Tong

ECCV 2020 · Point clouds › Point cloud classification · approach: Efficient point sampling architectures · 195 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
use 1 FC layer instead of 3 FC layers (ablation)point-wise MLP operator · S3DIS Area-5mIoU: 59.9 -> 62.9Improved
use 1 FC layer instead of 2 FC layers (ablation)point-wise MLP operator · S3DIS Area-5mIoU: 59.5 -> 62.9Improved
use MAX reduction instead of AVG or SUM (ablation)point-wise MLP operator · S3DIS Area-5mIoU: 44.7 (SUM), 55.1 (AVG) -> 62.9 (MAX)Improved
use 1 FC layer instead of 3 FC layers (ablation)adaptive weight operator · S3DIS Area-5mIoU: 58.5 -> 62.6Improved
use relative position input instead of edge feature input (ablation)adaptive weight operator · S3DIS Area-5mIoU: 46.6 -> 62.6Improved
add SoftMax normalization of aggregation weights (ablation)adaptive weight operator · S3DIS Area-5mIoU: 62.6 -> 55.9Hurt
use AVG reduction instead of SUM (ablation)adaptive weight operator · S3DIS Area-5mIoU: 61.7 -> 62.6Improved
add neighbor position encoding via PosPool over plain MAX poolingPosPool* · PartNet testmIoU: 47.4 -> 53.8Improved
use PosPool* instead of previous best PointCNNPosPool* · PartNet testmIoU: 46.4 -> 53.8Improved
use full relative position instead of angle or distance alone (ablation)PosPool variants · PartNet valmIoU: 43.4 -> 48.5Improved
use small 16x cheaper configuration (S) instead of regular for PosPool (ablation)PosPool · S3DIS Area-5mIoU: 66.5 -> 64.2Hurt
use Pseudo grid instead of point-wise MLP at regular sizeoperator choice · ModelNet40accuracy: 92.8 -> 93.0No clear effect

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