Table 2.
Precision, Recall, and F1-score values of the detection and identification results of different networks using different augmentation methods.
| Network | Augmentation method | Precision (%) | Recall (%) | F1-score |
|---|---|---|---|---|
| SSD | Rotation | 87 | 88.5 | 0.877 |
| Used in this study | 87 | 90 | 0.885 | |
| RetinaNet | Rotation | 94.75 | 88.5 | 0.915 |
| Used in this study | 95.5 | 90 | 0.927 | |
| Yolo V3 | Rotation | 86 | 89 | 0.875 |
| Used in this study | 75.25 | 91.5 | 0.825 | |
| Yolo V4 | Rotation | 97.75 | 79 | 0.874 |
| Used in this study | 96.75 | 80 | 0.876 | |
| Centernet | Rotation | 99 | 77.25 | 0.868 |
| Used in this study | 99.50 | 79.5 | 0.884 | |
| M2det | Rotation | 96.5 | 85 | 0.904 |
| Used in this study | 93.5 | 86 | 0.896 | |
| EfficientNet | Rotation | 97.25 | 84.25 | 0.898 |
| Used in this study | 98.25 | 84.25 | 0.907 | |
| AX-RetinaNet | Rotation | 93.75 | 92.25 | 0.930 |
| Used in this study | 96.75 | 94 | 0.954 |