Table 4.
Comparison of Recall on GC10-DET dataset. The proposed method performs the highest Recall values for seven defect categories. The bold helps to emphasize the highest data.
| Types | Recall | ||||
|---|---|---|---|---|---|
| SSD | Faster-RCNN | YOLO-V2 | YOLO-V3 | Proposed Method | |
| Pu | 0.964 | 0.964 | 0.857 | 0.964 | 0.965 |
| Wl | 1.000 | 0.623 | 0.869 | 0.869 | 0.967 |
| Cg | 0.968 | 0.968 | 0.936 | 0.871 | 0.969 |
| Ws | 0.696 | 0.696 | 0.674 | 0.609 | 0.739 |
| Os | 0.848 | 0.761 | 0.630 | 0.565 | 0.891 |
| Ss | 0.956 | 0.708 | 0.694 | 0.542 | 0.988 |
| In | 0.578 | 0.551 | 0.444 | 0.311 | 0.667 |
| Rp | 0.667 | 0.333 | 0.333 | 0.333 | 0.333 |
| Cr | 0.571 | 1.000 | 0.429 | 0.429 | 0.857 |
| Wf | 1.000 | 0.800 | 0.900 | 0.700 | 1.000 |