3D Point Cloud Classification on ScanObjectNN PB_50_RS 51
97.5AccuracyMOSTb=8
Evaluation results
| Method | Links | |
|---|---|---|
| MOSTb=8Backbone=PointGPT, 3D-designed (3D)=true, No inference overhead (NO)=true, Trainable Parameters (#P)=8.0M2025.03 | 97.5 | |
| MOSTb=16Backbone=PointGPT, 3D-designed (3D)=true, No inference overhead (NO)=true, Trainable Parameters (#P)=4.4M2025.03 | 97.26 | |
| MOSTb=32Backbone=PointGPT, 3D-designed (3D)=true, No inference overhead (NO)=true, Trainable Parameters (#P)=2.5M2025.03 | 97.19 | |
| Full FTBackbone=PointGPT, Trainable Parameters (#P)=360.5M2025.03 | 93.4 | |
| MOSTb=8Backbone=Point-MAE, 3D-designed (3D)=true, No inference overhead (NO)=true, Trainable Parameters (#P)=2.3M2025.03 | 92.92 | |
| MOSTb=16Backbone=Point-MAE, 3D-designed (3D)=true, No inference overhead (NO)=true, Trainable Parameters (#P)=1.3M2025.03 | 92.71 | |
| Full FTBackbone=Mamba3D, Trainable Parameters (#P)=16.9M2025.03 | 92.05 | |
| MOSTb=32Backbone=Point-MAE, 3D-designed (3D)=true, No inference overhead (NO)=true, Trainable Parameters (#P)=0.8M2025.03 | 91.95 | |
| Full FTBackbone=I2P-MAE, Trainable Parameters (#P)=15.3M2025.03 | 90.11 | |
| Full FTBackbone=ReCon, Trainable Parameters (#P)=22.1M2025.03 | 90.01 | |
| PointGSTBackbone=Point-MAE, 3D-designed (3D)=true, No inference overhead (NO)=false, Trainable Parameters (#P)=0.6M2025.03 | 89.3 | |
| PPTBackbone=Point-MAE, 3D-designed (3D)=true, No inference overhead (NO)=false, Trainable Parameters (#P)=1.1M2025.03 | 89 | |
| IDPTBackbone=Point-MAE, 3D-designed (3D)=true, No inference overhead (NO)=false, Trainable Parameters (#P)=1.7M2025.03 | 88.34 | |
| DAPTBackbone=Point-MAE, 3D-designed (3D)=true, No inference overhead (NO)=false, Trainable Parameters (#P)=1.1M2025.03 | 88.27 | |
| Full FTBackbone=Point-MAE, Trainable Parameters (#P)=22.1M2025.03 | 85.18 | |
| LoRABackbone=Point-MAE, 3D-designed (3D)=false, No inference overhead (NO)=true, Trainable Parameters (#P)=0.9M2025.03 | 82.76 | |
| BitFitBackbone=Point-MAE, 3D-designed (3D)=false, No inference overhead (NO)=true, Trainable Parameters (#P)=0.3M2025.03 | 82.62 | |
| VPTBackbone=Point-MAE, 3D-designed (3D)=false, No inference overhead (NO)=false, Trainable Parameters (#P)=0.4M2025.03 | 81.09 |