Research Article
Improved Faster R-CNN Based Surface Defect Detection Algorithm for Plates
Table 3
Scratch defect detection results.
| Defective texture background | Number of samples | Number of successful detections | Number of misses | Wrong number of checks | Positive inspection rate (%) |
| Same as the training set | 100 | 98 | ā | 1 | 98.50 | Unlike the training set | 40 | 36 | 3 | 1 | 91.60 | Total | 140 | 134 | 4 | 2 | 96.51 |
|
|