Table 6 Faster R-CNN performance on the Augmented PCB Defect dataset.
Method | mAP | \(D_1\) | \(D_2\) | \(D_3\) | \(D_4\) | \(D_5\) | \(D_6\) |
|---|---|---|---|---|---|---|---|
Faster R-CNN w/Resnet50 (baseline7a) | 97.6 | 96.4 | 97.3 | 97.4 | 97.1 | 98 | 99.3 |
Faster R-CNN w/Resnet101 (baseline7b) | 97.9 | 98 | 98 | 97.6 | 96.5 | 97.8 | 99.2 |
Faster R-CNN w/Resnet50+ ACASEM | 98.6 | 98.5 | 98.2 | 98.2 | 98 | 99.2 | 99.2 |
Faster R-CNN w/Resnet 101+ ACASEM | 98.5 | 98.9 | 97.6 | 98.3 | 97.8 | 99 | 99.3 |
