Point Cloud Classification Robustness on ModelNet40 (test)
0.59mCE (Error Rate)DGCNN + WOLFMix
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| DGCNN + WOLFMixTraining Strategy=Augmentations2024.01 | 0.59 | 0.989 | 0.715 | 0.698 | 0.575 | 0.285 | 0.415 | 0.451 | |
| DGCNN + RSMixTraining Strategy=Augmentations2024.01 | 0.745 | 1.319 | 0.873 | 0.653 | 0.589 | 0.281 | 0.629 | 0.87 | |
| PointMLSTraining Strategy=Architectures2024.01 | 0.784 | 1.181 | 0.652 | 0.565 | 0.913 | 0.522 | 0.756 | 0.902 | |
| DGCNN + PointWOLFTraining Strategy=Augmentations2024.01 | 0.814 | 0.926 | 0.864 | 0.988 | 0.874 | 0.807 | 0.764 | 0.479 | |
| RPCTraining Strategy=Architectures2024.01 | 0.863 | 0.84 | 0.892 | 0.492 | 0.797 | 0.929 | 1.011 | 1.079 | |
| GDANetTraining Strategy=Architectures2024.01 | 0.892 | 0.83 | 0.839 | 0.794 | 0.894 | 0.871 | 1.036 | 0.981 | |
| PCTTraining Strategy=Architectures2024.01 | 0.925 | 0.872 | 0.87 | 0.528 | 1 | 0.78 | 1.385 | 1.042 | |
| CurveNetTraining Strategy=Architectures2024.01 | 0.927 | 0.872 | 0.725 | 0.71 | 1.024 | 1.346 | 1 | 0.809 | |
| DGCNNTraining Strategy=Architectures2024.01 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | |
| PointNet2 + PointMixUpTraining Strategy=Augmentations2024.01 | 1.028 | 1.67 | 0.712 | 0.802 | 1.812 | 0.458 | 0.615 | 1.13 | |
| SimpleViewTraining Strategy=Architectures2024.01 | 1.047 | 0.872 | 0.715 | 1.242 | 1.357 | 0.983 | 0.844 | 1.316 | |
| DGCNN + OcCoTraining Strategy=Self-supervised Pre-training2024.01 | 1.047 | 1.606 | 0.652 | 0.903 | 1.039 | 1.444 | 0.847 | 0.837 | |
| PointNet2Training Strategy=Architectures2024.01 | 1.072 | 0.872 | 1.177 | 0.641 | 1.802 | 0.614 | 0.993 | 1.405 | |
| PAConvTraining Strategy=Architectures2024.01 | 1.104 | 0.904 | 1.465 | 1 | 1.005 | 1.085 | 1.298 | 0.967 | |
| RSCNNTraining Strategy=Architectures2024.01 | 1.13 | 1.074 | 1.171 | 0.806 | 1.517 | 0.712 | 1.153 | 1.479 | |
| Point-BERTTraining Strategy=Self-supervised Pre-training2024.01 | 1.248 | 0.936 | 1.259 | 0.69 | 1.15 | 1.932 | 1.44 | 1.326 | |
| PointNetTraining Strategy=Architectures2024.01 | 1.422 | 1.266 | 0.642 | 0.5 | 1.072 | 2.98 | 1.593 | 1.902 |