Classification on ScanObjectNN v2
87.7Overall Accuracy (OA)PointNeXt
Evaluation Results
| Method | Links | |
|---|---|---|
| PointNeXtArchitecture=MLP/CNN, Input points=1,024, Voting=false2026.05 | 87.7 | |
| PCM-TinyArchitecture=Mamba, Input points=1,024, Voting=false2026.05 | 86.9 | |
| Point-M2AEArchitecture=Transformer, Input points=1,024, Voting=false2026.05 | 86.4 | |
| PointMLPArchitecture=MLP/CNN, Input points=1,024, Voting=false2026.05 | 85.7 | |
| Point-MAEArchitecture=Transformer, Input points=1,024, Voting=false2026.05 | 85.2 | |
| PointMambaArchitecture=Mamba, Input points=1,024, Voting=false2026.05 | 84.9 | |
| RelFlexformer (w/ PCT)Architecture=Performer, Input points=1,024, Voting=false2026.05 | 84.5 | |
| MaskPointArchitecture=Transformer, Input points=1,024, Voting=false2026.05 | 84.3 | |
| RelFlexformer (w/ PCT) + PointRoPEArchitecture=Performer, Input points=1,024, Voting=false2026.05 | 84.3 | |
| PCTArchitecture=Transformer, Input points=1,024, Voting=false2026.05 | 84 | |
| Performer (w/ PCT)Architecture=Performer, Input points=1,024, Voting=false2026.05 | 83.2 | |
| Point-BERTArchitecture=Transformer, Input points=1,024, Voting=false2026.05 | 83.1 | |
| PointCNNArchitecture=MLP/CNN, Input points=1,024, Voting=false2026.05 | 78.5 | |
| DGCNNArchitecture=MLP/CNN, Input points=1,024, Voting=false2026.05 | 78.1 | |
| PointNet++Architecture=MLP/CNN, Input points=1,024, Voting=false2026.05 | 77.9 | |
| SpiderCNNArchitecture=MLP/CNN, Input points=1,024, Voting=false2026.05 | 73.7 | |
| PointNetArchitecture=MLP/CNN, Input points=1,024, Voting=false2026.05 | 68.2 |