List your company
ResearchFormerly WizWand

3D Point Cloud Classification on ModelNet40 (test)

94.5OAPointMLP+HyCoRe
92.83693.26893.794.132Jan 23, 2018Mar 22, 2019May 19, 2020Jul 16, 2021Sep 13, 2022Nov 10, 2023Jan 7, 2025
Updated 4mo ago

Evaluation results

MethodLinks
PointMLP+HyCoReTraining=regularized, voting=true2022.0994.5—91.9
PointMLPGroup Equivariant=false2022.1294.5——
PointMLPLearning Paradigm=Supervised Learning Only, #Params (M)=12.6, Input Data=1k Points, Voting=true2023.1294.5——
Point-FEMAELearning Paradigm=Single-Modal Self-Supervised Learning, #Params (M)=27.4, Input Data=1k Points, Voting=true2023.1294.5——
ReconLearning Paradigm=Cross-Modal Self-Supervised Learning, #Params (M)=44.3, Input Data=1k Points, Voting=true2023.1294.5——
PointMLP+HyCoReTraining=regularized2022.0994.3—91.7
CLIP2PointPoints=1,024, Views=102022.1094.2——
PointMLPInputs=xyz, #Points=10242023.1094.1——
PointMLPLearning Paradigm=Supervised Learning Only, #Params (M)=12.6, Input Data=1k Points, Voting=false2023.1294.1——
I2P-MAELearning Paradigm=Cross-Modal Self-Supervised Learning, #Params (M)=15.3, Input Data=1k Points, Voting=true2023.1294.1——
ReconLearning Paradigm=Cross-Modal Self-Supervised Learning, #Params (M)=44.3, Input Data=1k Points, Voting=false2023.1294.1——
P2P-HorNetLearning Paradigm=Supervised Learning Only, #Params (M)=195.8, Input Data=40 Images, Voting=false2023.1294——
Point-M2AELearning Paradigm=Single-Modal Self-Supervised Learning, #Params (M)=15.3, Input Data=1k Points, Voting=true2023.1294——
Point-FEMAELearning Paradigm=Single-Modal Self-Supervised Learning, #Params (M)=27.4, Input Data=1k Points, Voting=false2023.1294——
Joint-MAELearning Paradigm=Cross-Modal Self-Supervised Learning, Input Data=1k Points, Voting=false2023.1294——
P2P: HorNet-LPoints=4,096, Views=402022.1094——
SimpleView2022.0293.9——
PointHRInputs=xyz, #Points=10242023.1093.9——
SimpleViewLearning Paradigm=Supervised Learning Only, Input Data=6 Images, Voting=false2023.1293.9——
SFRLearning Paradigm=Supervised Learning Only, Input Data=12 Images, Voting=false2023.1293.9——
PointNet-CSRNumber of points=1024, Voting=false2025.0193.9—91.1
DOLPHIN (Best Idea)Backbone=PointNet, Mode=Maximum improvement2025.0193.991.1—
GDANet#params=0.93 M2020.1293.8——
CurveNet2022.0293.8——
CurveNetTraining=supervised2022.0993.8——
Point-MAEProtocol=FULL, Standard Transformer=false, #Point=1k P2022.1293.8——
MVTNLearning Paradigm=Supervised Learning Only, #Params (M)=11.2, Input Data=12 Images, Voting=false2023.1293.8——
MaskPointLearning Paradigm=Single-Modal Self-Supervised Learning, #Params (M)=22.1, Input Data=1k Points, Voting=false2023.1293.8——
Point-MAELearning Paradigm=Single-Modal Self-Supervised Learning, #Params (M)=22.1, Input Data=1k Points, Voting=true2023.1293.8——
GPSFormerNumber of points=1024, Voting=false2025.0193.8—91.8
GPSFormerBackbone=GPSFormer, Design=Human-designed2025.0193.891.8—
MVTNPoints=2,048, Views=122022.1093.8——
Transformer + Point-MAEPoints=1,024, Views=-2022.1093.8——
Point TransformerTraining=supervised2022.0993.7—90.6
PointMLPTraining=supervised, voting=true2022.0993.7—91.4
DGCNN+HyCoReTraining=regularized2022.0993.7—91
PointTransformerProtocol=Supervised Learning Only, Standard Transformer=false, #Point=1k P2022.1293.7——
ACTProtocol=FULL, Standard Transformer=false, #Point=1k P2022.1293.7——
PTv1Inputs=xyz, #Points=10242023.1093.7——
ACTLearning Paradigm=Cross-Modal Self-Supervised Learning, #Params (M)=22.1, Input Data=1k Points, Voting=true2023.1293.7——
I2P-MAELearning Paradigm=Cross-Modal Self-Supervised Learning, #Params (M)=15.3, Input Data=1k Points, Voting=false2023.1293.7——
PAConv2022.0293.6——
PA-DGCTraining=supervised2022.0993.6——
Inductive Bias-aided Transformer (IBT)Input Features=xyz, Point Count=1k2023.0493.691—
DGCNN (2048 points)graph_structure=dynamic, points=2048, k=402018.0193.590.7—
SO-NetAlignment Protocol=Pre-aligned, Input Points=PN50002018.0193.4——
EATinput=P, #points=1k2021.0593.4——
GDANet2022.0293.4——
PointMLPTraining=supervised2022.0993.4—91.2
RECONHierarchical=false, Evaluation protocol=Linear SVM, Voting=false2023.0293.4——
SO-NetInputs=xyz, #Points=50002023.1093.4——
Point-M2AELearning Paradigm=Single-Modal Self-Supervised Learning, #Params (M)=15.3, Input Data=1k Points, Voting=false2023.1293.4——
SimpleViewPoints=1,024, Views=62022.1093.4——
ARGNPL (#)=8, E=true, Params (M)=4.15, Search (Day)=0.22022.0593.33——
GS-NetInput=2k points, k=322019.1293.3——
3DCTNInput Features=xyz, normal, Point Count=1k2023.0493.391.2—
PointStackInput Features=xyz, Point Count=1k2023.0493.389.6—
DensePointnarrowness k=24, network depth L=11, #params=0.67M, #FLOPs/sample=651M2019.0993.2——
PointManifold (LLE Module)Module=LLE, Points=20482020.1093.290.2—
Centroid Transformer (Ours)MACs(G)=5.4, Params(M)=4.12021.0293.2——
PCTinput=P, #points=1k2021.0593.2——
PosPool*Inference Speed (instances/second)=27.62021.1093.2——
WOLFMixBackbone=DGCNN, Augmentation=WOLFMix2022.0293.2——
DGCNN+DCGLRTraining=finetuned2022.0993.2——
DensePointProtocol=Supervised Learning Only, #Point=1k P2022.1293.2——
PosPoolProtocol=Supervised Learning Only, #Point=5k P2022.1293.2——
PCTProtocol=Supervised Learning Only, Standard Transformer=false, #Point=1k P2022.1293.2——
Point-BERTProtocol=FULL, Standard Transformer=true, #Point=1k P2022.1293.2——
PointASNLInput Features=xyz, normal, Point Count=1k2023.0493.2——
PCTInput Features=xyz, Point Count=1k2023.0493.2——
PointNextInputs=xyz, #Points=10242023.1093.2——
PosPoolInputs=xyz, #Points=50002023.1093.2——
PCTInputs=xyz, #Points=10242023.1093.2——
Point-BERTLearning Paradigm=Single-Modal Self-Supervised Learning, #Params (M)=22.1, Input Data=1k Points, Voting=true2023.1293.2——
Point-MAELearning Paradigm=Single-Modal Self-Supervised Learning, #Params (M)=22.1, Input Data=1k Points, Voting=false2023.1293.2——
ACTLearning Paradigm=Cross-Modal Self-Supervised Learning, #Params (M)=22.1, Input Data=1k Points, Voting=false2023.1293.2——
CluRenderEvaluation Protocol=Linear Evaluation, Pre-training Dataset=ShapeNet, Backbone=Transformer2024.0493.2——
PointNeXtNumber of points=1024, Voting=false2025.0193.2—90.8
P2P: ConvNeXt-LPoints=4,096, Views=402022.1093.2——
Transformer + Point-BERTPoints=1,024, Views=-2022.1093.2——
PointManifold (MP Module)Module=MP, Points=20482020.1093.190.1—
DRNetinput type=coords, #points=1k2020.0593.1——
SepNet-W15MACs(G)=22.4, Params(M)=4.82021.0293.1——
Grid-GCNInference Speed (instances/second)=172.02021.1093.1——
DGCNN+STRLTraining=finetuned2022.0993.1——
DRNetProtocol=Supervised Learning Only, #Point=1k P2022.1293.1——
CloudTransformersInput Features=xyz, Point Count=1k2023.0493.190.8—
P2P: ResNet-101Points=4,096, Views=402022.1093.1——
Point2NodePoints=10242020.1093——
PointManifold (LLE Module)Module=LLE, Points=10242020.109390—
PointManifold (MP Module)Module=MP, Points=10242020.109390.1—
RSMixBackbone=DGCNN, Augmentation=RSMix2022.0293——
PointNet++Backbone=PointNet++, Augmentation=Standard2022.0293——
PointNet++2022.0293——
PCT2022.0293——
RPC2022.0293——
SGASL (#)=9, E=true, Params (M)=8.87, Search (Day)=0.192022.0592.93——
DGCNNgraph_structure=dynamic, points=10242018.0192.990.2—
GS-NetInput=1k points2019.1292.9——
KPConvPoints=10242020.1092.9——