Table 3 Research on the ablation of dual attention and feature fusion.
From: VM-UNet++ research on crack image segmentation based on improved VM-UNet
Attention block | Feature fusion module | Crack500 | Ozgenel | ||
|---|---|---|---|---|---|
mDS (%) | mIoU (%) | mDS (%) | mIoU (%) | ||
– | – | 70.3 | 56.0 | 85.7 | 79.4 |
√ | – | 73.0 | 57.5 | 89.8 | 81.5 |
– | √ | 73.1 | 57.6 | 89.8 | 81.5 |
√ | √ | 73.4 | 57.9 | 90.3 | 82.3 |
