List your company
ResearchFormerly WizWand

Point Cloud Classification on ScanObjectNN PB_T50_RS

90.63Overall AccuracyRecon
11.870832.317952.76573.2121Nov 26, 2022May 29, 2023Nov 29, 2023May 31, 2024Dec 1, 2024Jun 3, 2025Dec 5, 2025
Updated 6mo ago

Evaluation results

MethodLinks
ReconLearning Paradigm=Cross-Modal Self-Supervised Learning, #Params (M)=44.3, Input Data=2k Points2023.1290.63
Point-FEMAELearning Paradigm=Single-Modal Self-Supervised Learning, #Params (M)=27.4, Input Data=2k Points2023.1290.22
I2P-MAELearning Paradigm=Cross-Modal Self-Supervised Learning, #Params (M)=15.3, Input Data=2k Points2023.1290.11
PointGPT-BPretrain=GPT, #Params(M)=120.5, FLOPs(G)=36.2, Evaluation Setting=Original Paper2024.0589.6
P2P-HorNetLearning Paradigm=Supervised Learning Only, #Params (M)=195.8, Input Data=40 Images2023.1289.3
Point-MAE w/ LCMPretrain=MPM, #Params(M)=2.7, FLOPs(G)=1.32024.0588.87
Point-BERT w/ LCMPretrain=MPM, #Params(M)=3.1, FLOPs(G)=2.52024.0588.57
ACT w/ LCMPretrain=MPM, #Params(M)=3.1, FLOPs(G)=2.82024.0588.57
PointMLP + TAPBackbone=PointMLP, Pre-training=TAP2023.0788.5
Point-M2AE w/ LCMPretrain=MPM, #Params(M)=2.5, FLOPs(G)=6.72024.0588.38
Point-MAEPretrain=MPM, #Params(M)=22.1, FLOPs(G)=4.8, Evaluation Setting=Reproduced downstream setting2024.0588.27
ACTLearning Paradigm=Cross-Modal Self-Supervised Learning, #Params (M)=22.1, Input Data=2k Points2023.1288.21
ACTPretrain=MPM, #Params(M)=22.1, FLOPs(G)=4.8, Evaluation Setting=Original Paper2024.0588.21
PointMambaPretrain=MPM, #Params(M)=12.3, Evaluation Setting=Original Paper2024.0588.17
Point-M2AEPretrain=MPM, #Params(M)=12.9, FLOPs(G)=7.9, Evaluation Setting=Reproduced downstream setting2024.0588.06
Point-BERTPretrain=MPM, #Params(M)=22.1, FLOPs(G)=4.5, Evaluation Setting=Reproduced downstream setting2024.0587.91
SFRLearning Paradigm=Supervised Learning Only, Input Data=20 Images2023.1287.8
SFRPretrain=X, Evaluation Setting=Original Paper2024.0587.8
LCMPretrain=X, #Params(M)=2.7, FLOPs(G)=1.32024.0587.75
MaskPoint w/ LCMPretrain=MPM, #Params(M)=3.1, FLOPs(G)=2.52024.0587.75
MaskFeat3DBackbone=PointViT, Evaluation Protocol=fine-tuning2025.1287.7
MaskPointPretrain=MPM, #Params(M)=22.1, FLOPs(G)=4.5, Evaluation Setting=Reproduced downstream setting2024.0587.65
ACTPretrain=MPM, #Params(M)=22.1, FLOPs(G)=4.8, Evaluation Setting=Reproduced downstream setting2024.0587.52
PointMLPBackbone=PointMLP, Pre-training=None2023.0787.4
PointNet++ + TAPBackbone=PointNet++, Pre-training=TAP2023.0786.8
TransformerPretrain=X, #Params(M)=22.1, FLOPs(G)=4.8, Evaluation Setting=Reproduced downstream setting2024.0586.65
DGCNN + TAPBackbone=DGCNN, Pre-training=TAP2023.0786.6
Point-M2AELearning Paradigm=Single-Modal Self-Supervised Learning, #Params (M)=15.3, Input Data=2k Points2023.1286.43
Point-M2AEPretrain=MPM, #Params(M)=12.9, FLOPs(G)=7.9, Evaluation Setting=Original Paper2024.0586.43
PointNet++Backbone=PointNet++, Pre-training=None2023.0786.2
DGCNNBackbone=DGCNN, Pre-training=None2023.0786.1
Joint-MAELearning Paradigm=Cross-Modal Self-Supervised Learning, Input Data=2k Points2023.1286.07
baselinec=642025.1186.05
MaskSurfelBackbone=PointViT, Evaluation Protocol=fine-tuning2025.1285.7
TAP (Ours)Backbone=Standard Transformer, Pre-training=Generative2023.0785.67
Inter-MAEPretrain=MPM, #Params(M)=22.1, FLOPs(G)=4.8, Evaluation Setting=Original Paper2024.0585.4
PointMLPLearning Paradigm=Supervised Learning Only, #Params (M)=12.6, Input Data=1k Points2023.1285.2
PointMLPPretrain=X, #Params(M)=12.6, FLOPs(G)=31.4, Evaluation Setting=Original Paper2024.0585.2
PointMAEBackbone=PointViT, Evaluation Protocol=fine-tuning2025.1285.2
Point-MAEBackbone=Standard Transformer, Pre-training=Generative2023.0785.18
Point-MAELearning Paradigm=Single-Modal Self-Supervised Learning, #Params (M)=22.1, Input Data=2k Points2023.1285.18
Point-MAEPretrain=MPM, #Params(M)=22.1, FLOPs(G)=4.8, Evaluation Setting=Original Paper2024.0585.18
Point-MAE w/ DAPTPretrain=MPM, #Params(M)=22.7, FLOPs(G)=5.0, Evaluation Setting=Original Paper2024.0585.08
Point-MAE w/ IDPTPretrain=MPM, #Params(M)=23.3, FLOPs(G)=7.1, Evaluation Setting=Original Paper2024.0584.94
MaskPointBackbone=Standard Transformer, Pre-training=Generative2023.0784.3
MaskPointLearning Paradigm=Single-Modal Self-Supervised Learning, #Params (M)=22.1, Input Data=2k Points2023.1284.3
MaskPointPretrain=MPM, #Params(M)=22.1, FLOPs(G)=4.5, Evaluation Setting=Original Paper2024.0584.3
MaskDiscrBackbone=PointViT, Evaluation Protocol=fine-tuning2025.1284.3
PointBERTBackbone=PointViT, Evaluation Protocol=fine-tuning2025.1283.1
Point-BERTBackbone=Standard Transformer, Pre-training=Generative2023.0783.07
Point-BERTLearning Paradigm=Single-Modal Self-Supervised Learning, #Params (M)=22.1, Input Data=1k Points2023.1283.07
Point-BERTPretrain=MPM, #Params(M)=22.1, FLOPs(G)=4.5, Evaluation Setting=Original Paper2024.0583.07
MVTNLearning Paradigm=Supervised Learning Only, #Params (M)=11.2, Input Data=20 Images2023.1282.8
PointMambaPretrain=X, #Params(M)=12.3, Evaluation Setting=Original Paper2024.0582.48
TransformerPretrain=X, #Params(M)=22.1, FLOPs(G)=4.8, Evaluation Setting=Original Paper2024.0580.78
SimpleViewLearning Paradigm=Supervised Learning Only, Input Data=6 Images2023.1280.5
TetraSphereRobustness Category=Rotation-robust, K=4, Rotation Protocol=z/z2022.1179.2
TetraSphereRobustness Category=Rotation-robust, K=4, Rotation Protocol=z/SO(3)2022.1179.2
TetraSphereRobustness Category=Rotation-robust, K=2, Rotation Protocol=SO(3)/SO(3)2022.1179
TetraSphereRobustness Category=Rotation-robust, K=4, Rotation Protocol=SO(3)/SO(3)2022.1179
TetraSphereRobustness Category=Rotation-robust, K=8, Rotation Protocol=SO(3)/SO(3)2022.1179
TetraSphereRobustness Category=Rotation-robust, K=16, Rotation Protocol=SO(3)/SO(3)2022.1179
TetraSphereRobustness Category=Rotation-robust, K=2, Rotation Protocol=z/z2022.1178.9
TetraSphereRobustness Category=Rotation-robust, K=2, Rotation Protocol=z/SO(3)2022.1178.9
TetraSphereRobustness Category=Rotation-robust, K=16, Rotation Protocol=z/z2022.1178.8
TetraSphereRobustness Category=Rotation-robust, K=16, Rotation Protocol=z/SO(3)2022.1178.8
OcCoBackbone=Standard Transformer, Pre-training=Generative2023.0778.79
TetraSphereRobustness Category=Rotation-robust, K=1, Rotation Protocol=SO(3)/SO(3)2022.1178.7
TetraSphereRobustness Category=Rotation-robust, K=8, Rotation Protocol=z/z2022.1178.7
TetraSphereRobustness Category=Rotation-robust, K=8, Rotation Protocol=z/SO(3)2022.1178.7
PointCNNRobustness Category=Rotation-sensitive, Rotation Protocol=z/z2022.1178.5
VN-DGCNNRobustness Category=Rotation-robust, Rotation Protocol=SO(3)/SO(3)2022.1178.5
TetraSphereRobustness Category=Rotation-robust, K=1, Rotation Protocol=z/z2022.1178.5
TetraSphereRobustness Category=Rotation-robust, K=1, Rotation Protocol=z/SO(3)2022.1178.5
DGCNNLearning Paradigm=Supervised Learning Only, #Params (M)=1.8, Input Data=1k Points2023.1278.1
DGCNNRobustness Category=Rotation-sensitive, Rotation Protocol=z/z2022.1178.1
PointNet++Learning Paradigm=Supervised Learning Only, #Params (M)=1.5, Input Data=1k Points2023.1277.9
PointNet++Pretrain=X, #Params(M)=1.5, FLOPs(G)=1.7, Evaluation Setting=Original Paper2024.0577.9
VN-DGCNNRobustness Category=Rotation-robust, Rotation Protocol=z/z2022.1177.9
VN-DGCNNRobustness Category=Rotation-robust, Rotation Protocol=z/SO(3)2022.1177.9
Yu et al.Robustness Category=Rotation-robust, Rotation Protocol=SO(3)/SO(3)2022.1177.4
w/o pre-trainingBackbone=Standard Transformer, Pre-training=None2023.0777.24
PointViTBackbone=PointViT, Evaluation Protocol=from scratch2025.1277.2
Yu et al.Robustness Category=Rotation-robust, Rotation Protocol=z/z2022.1177.2
Yu et al.Robustness Category=Rotation-robust, Rotation Protocol=z/SO(3)2022.1177.2
Li et al.Robustness Category=Rotation-robust, Rotation Protocol=SO(3)/SO(3)2022.1174.9
Li et al.Robustness Category=Rotation-robust, Rotation Protocol=z/z2022.1174.6
Li et al.Robustness Category=Rotation-robust, Rotation Protocol=z/SO(3)2022.1174.6
3D-GFERobustness Category=Rotation-robust, Rotation Protocol=z/z2022.1173.5
3D-GFERobustness Category=Rotation-robust, Rotation Protocol=SO(3)/SO(3)2022.1173.5
3D-GFERobustness Category=Rotation-robust, Rotation Protocol=z/SO(3)2022.1172.7
PaRINetRobustness Category=Rotation-robust, Rotation Protocol=SO(3)/SO(3)2022.1172.2
PaRINetRobustness Category=Rotation-robust, Rotation Protocol=z/z2022.1171.6
PaRINetRobustness Category=Rotation-robust, Rotation Protocol=z/SO(3)2022.1171.6
PointNetLearning Paradigm=Supervised Learning Only, #Params (M)=3.5, Input Data=1k Points2023.1268
PointNetPretrain=X, #Params(M)=3.5, FLOPs(G)=0.5, Evaluation Setting=Original Paper2024.0568
DGCNNRobustness Category=Rotation-sensitive, Rotation Protocol=SO(3)/SO(3)2022.1163.4
PointCNNRobustness Category=Rotation-sensitive, Rotation Protocol=SO(3)/SO(3)2022.1151.8
DGCNNRobustness Category=Rotation-sensitive, Rotation Protocol=z/SO(3)2022.1116.1
PointCNNRobustness Category=Rotation-sensitive, Rotation Protocol=z/SO(3)2022.1114.9