Research Article
Detecting Gear Surface Defects Using Background-Weakening Method and Convolutional Neural Network
Table 5
S-YOLO network detection gear defect situation.
| Defect type | Number of defects in the test set | S-YOLO classification accuracy | YOLOv3 classification accuracy | Average confidence of each type of defect in S-YOLO | YOLOv3 average confidence of each type of defect |
| Broken tooth | 300 | 100% | 98.2% | 98.30% | 88.4% | Missing tooth | 300 | 100% | 98.7% | 97.05% | 84.9% | Surface scratch | 300 | 100% | 98.6% | 86.54% | 80.2% |
|
|