List your company
ResearchFormerly WizWand

3D Object Classification on ScanObjectNN

90.63AccuracyReCon
65.638872.126978.61585.1031Aug 16, 2024Nov 4, 2024Jan 23, 2025Apr 13, 2025Jul 2, 2025Sep 20, 2025Dec 9, 2025
Updated 5mo ago

Evaluation results

MethodLinks
ReConMasked Centers Leakage=true, Single/Cross-Modal=Cross, Pre-trained Model Needed=true, # Params=140.9, GFLOPS=18.2, Time (h)=44.42024.0890.63
PCP-MAEMasked Centers Leakage=false, Single/Cross-Modal=Single, Pre-trained Model Needed=false, # Params=29.5, GFLOPS=2.9, Time (h)=12.32024.0890.35
Point-FEMAEMasked Centers Leakage=true, Single/Cross-Modal=Single, Pre-trained Model Needed=false, # Params=41.5, GFLOPS=4.4, Time (h)=32.22024.0890.22
I2P-MAEMasked Centers Leakage=true, Single/Cross-Modal=Cross, Pre-trained Model Needed=true, # Params=74.9, GFLOPS=14.6, Time (h)=42.62024.0890.11
CSConFramework=Contrastive, #Params (M)=22.1 (0.76x), Local=true, Objective=high-level, Evaluation Protocol=MLP-32025.1289.8
ReConFramework=Generative, #Params (M)=140.9 (4.85x), Local=true, Objective=low-level, Evaluation Protocol=MLP-32025.1288.4
Point-PQAEFramework=Generative, #Params (M)=29.5 (1.02x), Local=true, Objective=low-level, Evaluation Protocol=MLP-32025.1288.3
ACTMasked Centers Leakage=true, Single/Cross-Modal=Cross, Pre-trained Model Needed=true, # Params=135.5, GFLOPS=23.0, Time (h)=34.82024.0888.21
PCP-MAEFramework=Generative, #Params (M)=29.5 (1.02x), Local=true, Objective=low-level, Evaluation Protocol=MLP-32025.1288.1
CSConFramework=Contrastive, #Params (M)=22.1 (0.76x), Local=true, Objective=high-level, Evaluation Protocol=LINEAR2025.1288
Point-FEMAEFramework=Generative, #Params (M)=41.5 (1.43x), Local=true, Objective=low-level, Evaluation Protocol=MLP-32025.1287.1
ReConFramework=Generative, #Params (M)=140.9 (4.85x), Local=true, Objective=low-level, Evaluation Protocol=LINEAR2025.1286.9
Point-PQAEFramework=Generative, #Params (M)=29.5 (1.02x), Local=true, Objective=low-level, Evaluation Protocol=LINEAR2025.1286.7
PCP-MAEFramework=Generative, #Params (M)=29.5 (1.02x), Local=true, Objective=low-level, Evaluation Protocol=LINEAR2025.1286.5
Point-M2AEMasked Centers Leakage=true, Single/Cross-Modal=Single, Pre-trained Model Needed=false, # Params=15.3, GFLOPS=3.2, Time (h)=19.12024.0886.43
Point-FEMAEFramework=Generative, #Params (M)=41.5 (1.43x), Local=true, Objective=low-level, Evaluation Protocol=LINEAR2025.1286.3
Point-MAEMasked Centers Leakage=true, Single/Cross-Modal=Single, Pre-trained Model Needed=false, # Params=29.0, GFLOPS=2.0, Time (h)=8.72024.0885.18
Spike-driven PointFormer-L + SVLArchitecture=SNN, Input=Point, Param (M)=22.1, Energy (mJ)=9.8, T × D=1 × 42025.0583.4
E-3DSNN-L + SVLArchitecture=SNN, Input=Voxel, Param (M)=17.7, Energy (mJ)=0.31, T × D=1 × 42025.0583
Point-MAEFramework=Generative, #Params (M)=29.0 (baseline), Local=true, Objective=low-level, Evaluation Protocol=MLP-32025.1282.3
Spike-driven PointFormer-SArchitecture=SNN, Input=Point, Param (M)=7.69, Energy (mJ)=5.1, T × D=1 × 42025.0582.1
CrossPointFramework=Contrastive, #Params (M)=27.7 (0.96x), Local=false, Objective=high-level, Evaluation Protocol=LINEAR2025.1281.7
Spike-driven PointFormer-LArchitecture=SNN, Input=Point, Param (M)=22.1, Energy (mJ)=9.4, T × D=1 × 42025.0581.7
PointformerArchitecture=ANN, Input=Point, Param (M)=4.91, Energy (mJ)=30.12025.0581.3
P2SResLNetArchitecture=SNN, Input=Point, Param (M)=14.3, T × D=4 × 12025.0581.2
E-3DSNN-S + SVLArchitecture=SNN, Input=Voxel, Param (M)=3.27, Energy (mJ)=0.02, T × D=1 × 42025.0580.9
E-3DSNN-LArchitecture=SNN, Input=Voxel, Param (M)=17.7, Energy (mJ)=0.26, T × D=1 × 42025.0580.2
Spike Point TransFormerArchitecture=SNN, Input=Point, Param (M)=9.6, Energy (mJ)=21.1, T × D=4 × 12025.0580.1
Point-MAEFramework=Generative, #Params (M)=29.0 (baseline), Local=true, Objective=low-level, Evaluation Protocol=LINEAR2025.1279.7
E-3DSNN-SArchitecture=SNN, Input=Voxel, Param (M)=3.27, Energy (mJ)=0.02, T × D=1 × 42025.0578.7
Spike PointNet + SVLArchitecture=SNN, Input=Point, Param (M)=3.47, Energy (mJ)=0.27, T × D=1 × 42025.0576.1
PointNet + ULIPArchitecture=ANN, Input=Point, Param (M)=3.47, Energy (mJ)=2.342025.0572.1
Spike PointNetArchitecture=SNN, Input=Point, Param (M)=3.47, Energy (mJ)=0.24, T × D=1 × 42025.0570
PointNetArchitecture=ANN, Input=Point, Param (M)=3.27, Energy (mJ)=2.022025.0568.2
SpikingPointNetArchitecture=SNN, Input=Point, Param (M)=3.47, Energy (mJ)=0.91, T × D=16 × 12025.0566.6