Table 3 The evaluation results (including IoU, DSC, and accuracy) of models trained by varying \(\gamma\) for FL loss (optimal \(\alpha\), dropout rate 0.75).
From: Wave based damage detection in solid structures using spatially asymmetric encoder–decoder network
\({\upgamma }\) | \({\upalpha }\) | Prec | Recall | IoU | DSC | Accu |
|---|---|---|---|---|---|---|
0 | 0.75 | 0.842 | 0.850 | 0.733 | 0.846 | 0.813 |
0.1 | 0.9 | 0.775 | 0.880 | 0.702 | 0.825 | 0.822 |
0.2 | 0.95 | 0.727 | 0.895 | 0.670 | 0.802 | 0.794 |
0.4 | 0.75 | 0.857 | 0.841 | 0.738 | 0.849 | 0.809 |
1 | 0.35 | 0.882 | 0.799 | 0.722 | 0.838 | 0.800 |
2 | 0.9 | 0.788 | 0.860 | 0.698 | 0.822 | 0.810 |
4 | 0.75 | 0.827 | 0.843 | 0.716 | 0.835 | 0.806 |
