Table 4 Prediction results of BWO–VMD–HHT datasets.
From: Defect monitoring method for Al-CFRTP UFSW based on BWO–VMD–HHT and ResNet
Classification method | Precision ↑ | Recall ↑ | F1-Score ↑ | GFLOPs↓ |
|---|---|---|---|---|
ResNet18-Attention (with − 10 dB noise dataset) | 0.611 | 0.614 | 0.612 | 1.84 |
Multi-SVM | 0.703 | 0.704 | 0.703 | 0.52 |
Hartl33 (CNN) | 0.792 | – | – | 1.20 |
RBF | 0.803 | 0.804 | 0.802 | – |
BP-Net | 0.812 | 0.813 | 0.812 | 0.21 |
Li26 (Res-GCM) | 0.840 | – | – | 3.24 |
Rabe34 (BiLSTM) | 0.865 | – | – | 2.02 |
ResNet18 | 0.886 | 0.878 | 0.876 | 1.82 |
ResNet18-attention (with − 2 dB noise dataset) | 0.891 | 0.884 | 0.877 | 1.84 |
ResNet18-attention (this paper) | 0.914 | 0.915 | 0.914 | 1.84 |
