List your company
ResearchFormerly WizWand

3D object classification on ScanObjectNN OBJ_ONLY

94.2Overall AccuracyMVNet-L
20.56839.68458.877.916Nov 26, 2020Oct 27, 2021Sep 28, 2022Aug 29, 2023Jul 30, 2024Jun 30, 2025Jun 1, 2026
Updated 5mo ago

Evaluation results

MethodLinks
MVNet-LPre-training Dataset=Objaverse, Feature Dimension=7682025.1294.2
3D-JEPAPre-training Epochs=300, Training Paradigm=SSRL2024.0993.63
PointSD2026.0693.6
ReCon#P (M)=43.6, #F (G)=5.3, Training Paradigm=with Pretrained Cross-Modal Teacher Representation Learning2023.0793.29
Point-DAE2026.0693.1
3D-JEPAPre-training Epochs=150, Training Paradigm=SSRL2024.0992.77
PRISMcombination=with Point-MAE2026.0692.7
MVTN2020.1192.3
PRISM2026.0692.1
MaskFeat3DBackbone=PointNeXt2025.1292
VPP w/ vot.#P (M)=22.1, #F (G)=4.8, Training Paradigm=with Self-Supervised Representation Learning, Voting Strategy=Yes2023.0791.91
ACT#P (M)=22.1, #F (G)=4.8, Training Paradigm=with Pretrained Cross-Modal Teacher Representation Learning2023.0791.91
ACTPre-training Epochs=300, Training Paradigm=SSRL2024.0991.91
PointDif2026.0691.9
IAE (M2AE)#Params(M)=15.3, GFLOPS=3.6, Evaluation Protocol=FULL SSL2025.1291.6
I2P-MAE#P (M)=12.9, #F (G)=3.6, Training Paradigm=with Pretrained Cross-Modal Teacher Representation Learning2023.0791.57
VPP w/o vot.#P (M)=22.1, #F (G)=4.8, Training Paradigm=with Self-Supervised Representation Learning, Voting Strategy=No2023.0791.56
IAE (M2AE)#Params(M)=15.3, GFLOPS=3.6, Evaluation Protocol=FULL SSL, pre-training mesh=w/o mesh2025.1291.2
PointNeXt2025.1291
Point2VecPre-training Epochs=300, Training Paradigm=SSRL2024.0990.4
MVNet-BPre-training Dataset=Objaverse2025.1290.1
MaskFeat3DBackbone=PointViT2025.1290
MVNet-BPre-training Dataset=ShapeNet2025.1289.7
ULIPPre-training Dataset=ShapeNet2025.1289.4
MaskDiscrPre-training Dataset=ShapeNet2025.1289.3
MaskSurfelPre-training Dataset=ShapeNet2025.1289.2
IAE (DGCNN)#Params(M)=1.8, GFLOPS=2.4, Evaluation Protocol=FULL SSL2025.1289
Point-M2AE#P (M)=14.8, #F (G)=3.6, Training Paradigm=with Self-Supervised Representation Learning2023.0788.81
Point-M2AEPre-training Epochs=300, Training Paradigm=SSRL2024.0988.81
Point-M2AE#Params(M)=15.3, GFLOPS=3.6, Evaluation Protocol=FULL SSL2025.1288.8
Point-M2AE2026.0688.8
3D-OAEPre-training Epochs=300, Training Paradigm=SSRL2024.0988.64
Point-MAE#Params(M)=22.1, GFLOPS=4.8, Evaluation Protocol=FULL SSL2025.1288.3
PointMAEPre-training Dataset=ShapeNet2025.1288.3
Point-MAE2026.0688.3
Point-MAE#P (M)=22.1, #F (G)=4.8, Training Paradigm=with Self-Supervised Representation Learning2023.0788.29
Point-MAEPre-training Epochs=300, Training Paradigm=SSRL2024.0988.29
PointMLP2025.1288.2
PTv3training=from scratch2026.0688.2
Point-BERT#P (M)=22.1, #F (G)=4.8, Training Paradigm=with Self-Supervised Representation Learning2023.0788.12
Point-BERTPre-training Epochs=300, Training Paradigm=SSRL2024.0988.12
Point-BERT#Params(M)=22.1, GFLOPS=4.8, Evaluation Protocol=FULL SSL2025.1288.1
PointBERTPre-training Dataset=ShapeNet2025.1288.1
TAMMPre-Training Dataset=Ensembled, Evaluation Protocol=Linear Probing2024.0288
OcCo#Params(M)=1.8, GFLOPS=2.4, Evaluation Protocol=FULL SSL2025.1287.5
Point-BERT2026.0687.4
MaskFeat3DBackbone=MinkowskiNet2025.1287
DGCNN#P (M)=1.8, #F (G)=2.4, Training Paradigm=Supervised Learning Only2023.0786.2
DGCNNPre-training Epochs=N/A, Training Paradigm=Supervised Learning Only2024.0986.2
DGCNN#Params(M)=1.8, GFLOPS=2.4, Evaluation Protocol=FULL SSL2025.1286.2
DGCNN2025.1286.2
DGCNN2020.1186.2
MinkowskiNet2025.1286.1
PointCNN#P (M)=0.6, Training Paradigm=Supervised Learning Only2023.0785.5
PointCNN#Params(M)=0.6, Evaluation Protocol=Supervised2025.1285.5
PointCNN2025.1285.5
PointCNN2020.1185.5
OpenShapePre-Training Dataset=Ensembled, Evaluation Protocol=Linear Probing2024.0285.4
PointNet++#P (M)=1.5, #F (G)=1.7, Training Paradigm=Supervised Learning Only2023.0784.3
PointNet++Pre-training Epochs=N/A, Training Paradigm=Supervised Learning Only2024.0984.3
PointNet++#Params(M)=1.5, GFLOPS=1.7, Evaluation Protocol=Supervised2025.1284.3
PointNet++2025.1284.3
PointNet ++2020.1184.3
KPConvEvaluation Protocol=Supervised2025.1284.1
Transformer#Params(M)=22.1, GFLOPS=4.8, Evaluation Protocol=FULL SSL2025.1284.1
Transformer#P (M)=22.1, #F (G)=4.8, Training Paradigm=with Self-Supervised Representation Learning2023.0784.06
TransformerPre-training Epochs=300, Training Paradigm=SSRL2024.0984.06
TAMMPre-Training Dataset=ShapeNet, Evaluation Protocol=Linear Probing2024.0281.1
TransformerEvaluation Protocol=from scratch2025.1280.6
SpiderCNN2020.1179.5
PointNet#P (M)=3.5, #F (G)=0.5, Training Paradigm=Supervised Learning Only2023.0779.2
PointNetPre-training Epochs=N/A, Training Paradigm=Supervised Learning Only2024.0979.2
PointNet#Params(M)=3.5, GFLOPS=0.5, Evaluation Protocol=Supervised2025.1279.2
PointNet2025.1279.2
PointNet2020.1179.2
OpenShapePre-Training Dataset=ShapeNet, Evaluation Protocol=Linear Probing2024.0278.5
ULIPPre-Training Dataset=ShapeNet, Evaluation Protocol=Linear Probing2024.0275.4
3DMFV2020.1173.8
P3TCross-dataset generalization=true, Source dataset=Objaverse-LVIS2026.0452.8
Point-PRCCross-dataset generalization=true, Source dataset=Objaverse-LVIS2026.0445.1
PPTCross-dataset generalization=true, Source dataset=Objaverse-LVIS2026.0439.8
ULIP-2 (ZS)Zero-shot setting=true, Cross-dataset generalization=true, Source dataset=Objaverse-LVIS2026.0432.2
ULIP-2Cross-dataset generalization=true, Source dataset=Objaverse-LVIS2026.0423.4